summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
blob: f2805ba74c80be6d9fd51701f82b8b0e9feb0f54 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
// SPDX-License-Identifier: GPL-2.0 OR MIT
/*
 * Copyright 2020-2021 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 */

#include <linux/types.h>
#include <linux/sched/task.h>
#include "amdgpu_sync.h"
#include "amdgpu_object.h"
#include "amdgpu_vm.h"
#include "amdgpu_mn.h"
#include "amdgpu.h"
#include "amdgpu_xgmi.h"
#include "kfd_priv.h"
#include "kfd_svm.h"
#include "kfd_migrate.h"

#ifdef dev_fmt
#undef dev_fmt
#endif
#define dev_fmt(fmt) "kfd_svm: %s: " fmt, __func__

#define AMDGPU_SVM_RANGE_RESTORE_DELAY_MS 1

/* Long enough to ensure no retry fault comes after svm range is restored and
 * page table is updated.
 */
#define AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING	2000

static void svm_range_evict_svm_bo_worker(struct work_struct *work);
static bool
svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni,
				    const struct mmu_notifier_range *range,
				    unsigned long cur_seq);
static int
svm_range_check_vm(struct kfd_process *p, uint64_t start, uint64_t last,
		   uint64_t *bo_s, uint64_t *bo_l);
static const struct mmu_interval_notifier_ops svm_range_mn_ops = {
	.invalidate = svm_range_cpu_invalidate_pagetables,
};

/**
 * svm_range_unlink - unlink svm_range from lists and interval tree
 * @prange: svm range structure to be removed
 *
 * Remove the svm_range from the svms and svm_bo lists and the svms
 * interval tree.
 *
 * Context: The caller must hold svms->lock
 */
static void svm_range_unlink(struct svm_range *prange)
{
	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms,
		 prange, prange->start, prange->last);

	if (prange->svm_bo) {
		spin_lock(&prange->svm_bo->list_lock);
		list_del(&prange->svm_bo_list);
		spin_unlock(&prange->svm_bo->list_lock);
	}

	list_del(&prange->list);
	if (prange->it_node.start != 0 && prange->it_node.last != 0)
		interval_tree_remove(&prange->it_node, &prange->svms->objects);
}

static void
svm_range_add_notifier_locked(struct mm_struct *mm, struct svm_range *prange)
{
	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms,
		 prange, prange->start, prange->last);

	mmu_interval_notifier_insert_locked(&prange->notifier, mm,
				     prange->start << PAGE_SHIFT,
				     prange->npages << PAGE_SHIFT,
				     &svm_range_mn_ops);
}

/**
 * svm_range_add_to_svms - add svm range to svms
 * @prange: svm range structure to be added
 *
 * Add the svm range to svms interval tree and link list
 *
 * Context: The caller must hold svms->lock
 */
static void svm_range_add_to_svms(struct svm_range *prange)
{
	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms,
		 prange, prange->start, prange->last);

	list_move_tail(&prange->list, &prange->svms->list);
	prange->it_node.start = prange->start;
	prange->it_node.last = prange->last;
	interval_tree_insert(&prange->it_node, &prange->svms->objects);
}

static void svm_range_remove_notifier(struct svm_range *prange)
{
	pr_debug("remove notifier svms 0x%p prange 0x%p [0x%lx 0x%lx]\n",
		 prange->svms, prange,
		 prange->notifier.interval_tree.start >> PAGE_SHIFT,
		 prange->notifier.interval_tree.last >> PAGE_SHIFT);

	if (prange->notifier.interval_tree.start != 0 &&
	    prange->notifier.interval_tree.last != 0)
		mmu_interval_notifier_remove(&prange->notifier);
}

static bool
svm_is_valid_dma_mapping_addr(struct device *dev, dma_addr_t dma_addr)
{
	return dma_addr && !dma_mapping_error(dev, dma_addr) &&
	       !(dma_addr & SVM_RANGE_VRAM_DOMAIN);
}

static int
svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
		      unsigned long offset, unsigned long npages,
		      unsigned long *hmm_pfns, uint32_t gpuidx)
{
	enum dma_data_direction dir = DMA_BIDIRECTIONAL;
	dma_addr_t *addr = prange->dma_addr[gpuidx];
	struct device *dev = adev->dev;
	struct page *page;
	int i, r;

	if (!addr) {
		addr = kvmalloc_array(prange->npages, sizeof(*addr),
				      GFP_KERNEL | __GFP_ZERO);
		if (!addr)
			return -ENOMEM;
		prange->dma_addr[gpuidx] = addr;
	}

	addr += offset;
	for (i = 0; i < npages; i++) {
		if (svm_is_valid_dma_mapping_addr(dev, addr[i]))
			dma_unmap_page(dev, addr[i], PAGE_SIZE, dir);

		page = hmm_pfn_to_page(hmm_pfns[i]);
		if (is_zone_device_page(page)) {
			struct amdgpu_device *bo_adev =
					amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);

			addr[i] = (hmm_pfns[i] << PAGE_SHIFT) +
				   bo_adev->vm_manager.vram_base_offset -
				   bo_adev->kfd.dev->pgmap.range.start;
			addr[i] |= SVM_RANGE_VRAM_DOMAIN;
			pr_debug_ratelimited("vram address: 0x%llx\n", addr[i]);
			continue;
		}
		addr[i] = dma_map_page(dev, page, 0, PAGE_SIZE, dir);
		r = dma_mapping_error(dev, addr[i]);
		if (r) {
			dev_err(dev, "failed %d dma_map_page\n", r);
			return r;
		}
		pr_debug_ratelimited("dma mapping 0x%llx for page addr 0x%lx\n",
				     addr[i] >> PAGE_SHIFT, page_to_pfn(page));
	}
	return 0;
}

static int
svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
		  unsigned long offset, unsigned long npages,
		  unsigned long *hmm_pfns)
{
	struct kfd_process *p;
	uint32_t gpuidx;
	int r;

	p = container_of(prange->svms, struct kfd_process, svms);

	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
		struct kfd_process_device *pdd;

		pr_debug("mapping to gpu idx 0x%x\n", gpuidx);
		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
		if (!pdd) {
			pr_debug("failed to find device idx %d\n", gpuidx);
			return -EINVAL;
		}

		r = svm_range_dma_map_dev(pdd->dev->adev, prange, offset, npages,
					  hmm_pfns, gpuidx);
		if (r)
			break;
	}

	return r;
}

void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr,
			 unsigned long offset, unsigned long npages)
{
	enum dma_data_direction dir = DMA_BIDIRECTIONAL;
	int i;

	if (!dma_addr)
		return;

	for (i = offset; i < offset + npages; i++) {
		if (!svm_is_valid_dma_mapping_addr(dev, dma_addr[i]))
			continue;
		pr_debug_ratelimited("unmap 0x%llx\n", dma_addr[i] >> PAGE_SHIFT);
		dma_unmap_page(dev, dma_addr[i], PAGE_SIZE, dir);
		dma_addr[i] = 0;
	}
}

void svm_range_free_dma_mappings(struct svm_range *prange)
{
	struct kfd_process_device *pdd;
	dma_addr_t *dma_addr;
	struct device *dev;
	struct kfd_process *p;
	uint32_t gpuidx;

	p = container_of(prange->svms, struct kfd_process, svms);

	for (gpuidx = 0; gpuidx < MAX_GPU_INSTANCE; gpuidx++) {
		dma_addr = prange->dma_addr[gpuidx];
		if (!dma_addr)
			continue;

		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
		if (!pdd) {
			pr_debug("failed to find device idx %d\n", gpuidx);
			continue;
		}
		dev = &pdd->dev->pdev->dev;
		svm_range_dma_unmap(dev, dma_addr, 0, prange->npages);
		kvfree(dma_addr);
		prange->dma_addr[gpuidx] = NULL;
	}
}

static void svm_range_free(struct svm_range *prange)
{
	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms, prange,
		 prange->start, prange->last);

	svm_range_vram_node_free(prange);
	svm_range_free_dma_mappings(prange);
	mutex_destroy(&prange->lock);
	mutex_destroy(&prange->migrate_mutex);
	kfree(prange);
}

static void
svm_range_set_default_attributes(int32_t *location, int32_t *prefetch_loc,
				 uint8_t *granularity, uint32_t *flags)
{
	*location = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
	*prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
	*granularity = 9;
	*flags =
		KFD_IOCTL_SVM_FLAG_HOST_ACCESS | KFD_IOCTL_SVM_FLAG_COHERENT;
}

static struct
svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start,
			 uint64_t last)
{
	uint64_t size = last - start + 1;
	struct svm_range *prange;
	struct kfd_process *p;

	prange = kzalloc(sizeof(*prange), GFP_KERNEL);
	if (!prange)
		return NULL;
	prange->npages = size;
	prange->svms = svms;
	prange->start = start;
	prange->last = last;
	INIT_LIST_HEAD(&prange->list);
	INIT_LIST_HEAD(&prange->update_list);
	INIT_LIST_HEAD(&prange->svm_bo_list);
	INIT_LIST_HEAD(&prange->deferred_list);
	INIT_LIST_HEAD(&prange->child_list);
	atomic_set(&prange->invalid, 0);
	prange->validate_timestamp = 0;
	mutex_init(&prange->migrate_mutex);
	mutex_init(&prange->lock);

	p = container_of(svms, struct kfd_process, svms);
	if (p->xnack_enabled)
		bitmap_copy(prange->bitmap_access, svms->bitmap_supported,
			    MAX_GPU_INSTANCE);

	svm_range_set_default_attributes(&prange->preferred_loc,
					 &prange->prefetch_loc,
					 &prange->granularity, &prange->flags);

	pr_debug("svms 0x%p [0x%llx 0x%llx]\n", svms, start, last);

	return prange;
}

static bool svm_bo_ref_unless_zero(struct svm_range_bo *svm_bo)
{
	if (!svm_bo || !kref_get_unless_zero(&svm_bo->kref))
		return false;

	return true;
}

static void svm_range_bo_release(struct kref *kref)
{
	struct svm_range_bo *svm_bo;

	svm_bo = container_of(kref, struct svm_range_bo, kref);
	pr_debug("svm_bo 0x%p\n", svm_bo);

	spin_lock(&svm_bo->list_lock);
	while (!list_empty(&svm_bo->range_list)) {
		struct svm_range *prange =
				list_first_entry(&svm_bo->range_list,
						struct svm_range, svm_bo_list);
		/* list_del_init tells a concurrent svm_range_vram_node_new when
		 * it's safe to reuse the svm_bo pointer and svm_bo_list head.
		 */
		list_del_init(&prange->svm_bo_list);
		spin_unlock(&svm_bo->list_lock);

		pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms,
			 prange->start, prange->last);
		mutex_lock(&prange->lock);
		prange->svm_bo = NULL;
		mutex_unlock(&prange->lock);

		spin_lock(&svm_bo->list_lock);
	}
	spin_unlock(&svm_bo->list_lock);
	if (!dma_fence_is_signaled(&svm_bo->eviction_fence->base)) {
		/* We're not in the eviction worker.
		 * Signal the fence and synchronize with any
		 * pending eviction work.
		 */
		dma_fence_signal(&svm_bo->eviction_fence->base);
		cancel_work_sync(&svm_bo->eviction_work);
	}
	dma_fence_put(&svm_bo->eviction_fence->base);
	amdgpu_bo_unref(&svm_bo->bo);
	kfree(svm_bo);
}

static void svm_range_bo_wq_release(struct work_struct *work)
{
	struct svm_range_bo *svm_bo;

	svm_bo = container_of(work, struct svm_range_bo, release_work);
	svm_range_bo_release(&svm_bo->kref);
}

static void svm_range_bo_release_async(struct kref *kref)
{
	struct svm_range_bo *svm_bo;

	svm_bo = container_of(kref, struct svm_range_bo, kref);
	pr_debug("svm_bo 0x%p\n", svm_bo);
	INIT_WORK(&svm_bo->release_work, svm_range_bo_wq_release);
	schedule_work(&svm_bo->release_work);
}

void svm_range_bo_unref_async(struct svm_range_bo *svm_bo)
{
	kref_put(&svm_bo->kref, svm_range_bo_release_async);
}

static void svm_range_bo_unref(struct svm_range_bo *svm_bo)
{
	if (svm_bo)
		kref_put(&svm_bo->kref, svm_range_bo_release);
}

static bool
svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange)
{
	struct amdgpu_device *bo_adev;

	mutex_lock(&prange->lock);
	if (!prange->svm_bo) {
		mutex_unlock(&prange->lock);
		return false;
	}
	if (prange->ttm_res) {
		/* We still have a reference, all is well */
		mutex_unlock(&prange->lock);
		return true;
	}
	if (svm_bo_ref_unless_zero(prange->svm_bo)) {
		/*
		 * Migrate from GPU to GPU, remove range from source bo_adev
		 * svm_bo range list, and return false to allocate svm_bo from
		 * destination adev.
		 */
		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
		if (bo_adev != adev) {
			mutex_unlock(&prange->lock);

			spin_lock(&prange->svm_bo->list_lock);
			list_del_init(&prange->svm_bo_list);
			spin_unlock(&prange->svm_bo->list_lock);

			svm_range_bo_unref(prange->svm_bo);
			return false;
		}
		if (READ_ONCE(prange->svm_bo->evicting)) {
			struct dma_fence *f;
			struct svm_range_bo *svm_bo;
			/* The BO is getting evicted,
			 * we need to get a new one
			 */
			mutex_unlock(&prange->lock);
			svm_bo = prange->svm_bo;
			f = dma_fence_get(&svm_bo->eviction_fence->base);
			svm_range_bo_unref(prange->svm_bo);
			/* wait for the fence to avoid long spin-loop
			 * at list_empty_careful
			 */
			dma_fence_wait(f, false);
			dma_fence_put(f);
		} else {
			/* The BO was still around and we got
			 * a new reference to it
			 */
			mutex_unlock(&prange->lock);
			pr_debug("reuse old bo svms 0x%p [0x%lx 0x%lx]\n",
				 prange->svms, prange->start, prange->last);

			prange->ttm_res = prange->svm_bo->bo->tbo.resource;
			return true;
		}

	} else {
		mutex_unlock(&prange->lock);
	}

	/* We need a new svm_bo. Spin-loop to wait for concurrent
	 * svm_range_bo_release to finish removing this range from
	 * its range list. After this, it is safe to reuse the
	 * svm_bo pointer and svm_bo_list head.
	 */
	while (!list_empty_careful(&prange->svm_bo_list))
		;

	return false;
}

static struct svm_range_bo *svm_range_bo_new(void)
{
	struct svm_range_bo *svm_bo;

	svm_bo = kzalloc(sizeof(*svm_bo), GFP_KERNEL);
	if (!svm_bo)
		return NULL;

	kref_init(&svm_bo->kref);
	INIT_LIST_HEAD(&svm_bo->range_list);
	spin_lock_init(&svm_bo->list_lock);

	return svm_bo;
}

int
svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
			bool clear)
{
	struct amdgpu_bo_param bp;
	struct svm_range_bo *svm_bo;
	struct amdgpu_bo_user *ubo;
	struct amdgpu_bo *bo;
	struct kfd_process *p;
	struct mm_struct *mm;
	int r;

	p = container_of(prange->svms, struct kfd_process, svms);
	pr_debug("pasid: %x svms 0x%p [0x%lx 0x%lx]\n", p->pasid, prange->svms,
		 prange->start, prange->last);

	if (svm_range_validate_svm_bo(adev, prange))
		return 0;

	svm_bo = svm_range_bo_new();
	if (!svm_bo) {
		pr_debug("failed to alloc svm bo\n");
		return -ENOMEM;
	}
	mm = get_task_mm(p->lead_thread);
	if (!mm) {
		pr_debug("failed to get mm\n");
		kfree(svm_bo);
		return -ESRCH;
	}
	svm_bo->svms = prange->svms;
	svm_bo->eviction_fence =
		amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1),
					   mm,
					   svm_bo);
	mmput(mm);
	INIT_WORK(&svm_bo->eviction_work, svm_range_evict_svm_bo_worker);
	svm_bo->evicting = 0;
	memset(&bp, 0, sizeof(bp));
	bp.size = prange->npages * PAGE_SIZE;
	bp.byte_align = PAGE_SIZE;
	bp.domain = AMDGPU_GEM_DOMAIN_VRAM;
	bp.flags = AMDGPU_GEM_CREATE_NO_CPU_ACCESS;
	bp.flags |= clear ? AMDGPU_GEM_CREATE_VRAM_CLEARED : 0;
	bp.flags |= AMDGPU_AMDKFD_CREATE_SVM_BO;
	bp.type = ttm_bo_type_device;
	bp.resv = NULL;

	r = amdgpu_bo_create_user(adev, &bp, &ubo);
	if (r) {
		pr_debug("failed %d to create bo\n", r);
		goto create_bo_failed;
	}
	bo = &ubo->bo;
	r = amdgpu_bo_reserve(bo, true);
	if (r) {
		pr_debug("failed %d to reserve bo\n", r);
		goto reserve_bo_failed;
	}

	r = dma_resv_reserve_shared(bo->tbo.base.resv, 1);
	if (r) {
		pr_debug("failed %d to reserve bo\n", r);
		amdgpu_bo_unreserve(bo);
		goto reserve_bo_failed;
	}
	amdgpu_bo_fence(bo, &svm_bo->eviction_fence->base, true);

	amdgpu_bo_unreserve(bo);

	svm_bo->bo = bo;
	prange->svm_bo = svm_bo;
	prange->ttm_res = bo->tbo.resource;
	prange->offset = 0;

	spin_lock(&svm_bo->list_lock);
	list_add(&prange->svm_bo_list, &svm_bo->range_list);
	spin_unlock(&svm_bo->list_lock);

	return 0;

reserve_bo_failed:
	amdgpu_bo_unref(&bo);
create_bo_failed:
	dma_fence_put(&svm_bo->eviction_fence->base);
	kfree(svm_bo);
	prange->ttm_res = NULL;

	return r;
}

void svm_range_vram_node_free(struct svm_range *prange)
{
	svm_range_bo_unref(prange->svm_bo);
	prange->ttm_res = NULL;
}

struct amdgpu_device *
svm_range_get_adev_by_id(struct svm_range *prange, uint32_t gpu_id)
{
	struct kfd_process_device *pdd;
	struct kfd_process *p;
	int32_t gpu_idx;

	p = container_of(prange->svms, struct kfd_process, svms);

	gpu_idx = kfd_process_gpuidx_from_gpuid(p, gpu_id);
	if (gpu_idx < 0) {
		pr_debug("failed to get device by id 0x%x\n", gpu_id);
		return NULL;
	}
	pdd = kfd_process_device_from_gpuidx(p, gpu_idx);
	if (!pdd) {
		pr_debug("failed to get device by idx 0x%x\n", gpu_idx);
		return NULL;
	}

	return pdd->dev->adev;
}

struct kfd_process_device *
svm_range_get_pdd_by_adev(struct svm_range *prange, struct amdgpu_device *adev)
{
	struct kfd_process *p;
	int32_t gpu_idx, gpuid;
	int r;

	p = container_of(prange->svms, struct kfd_process, svms);

	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpu_idx);
	if (r) {
		pr_debug("failed to get device id by adev %p\n", adev);
		return NULL;
	}

	return kfd_process_device_from_gpuidx(p, gpu_idx);
}

static int svm_range_bo_validate(void *param, struct amdgpu_bo *bo)
{
	struct ttm_operation_ctx ctx = { false, false };

	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM);

	return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
}

static int
svm_range_check_attr(struct kfd_process *p,
		     uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs)
{
	uint32_t i;

	for (i = 0; i < nattr; i++) {
		uint32_t val = attrs[i].value;
		int gpuidx = MAX_GPU_INSTANCE;

		switch (attrs[i].type) {
		case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC:
			if (val != KFD_IOCTL_SVM_LOCATION_SYSMEM &&
			    val != KFD_IOCTL_SVM_LOCATION_UNDEFINED)
				gpuidx = kfd_process_gpuidx_from_gpuid(p, val);
			break;
		case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC:
			if (val != KFD_IOCTL_SVM_LOCATION_SYSMEM)
				gpuidx = kfd_process_gpuidx_from_gpuid(p, val);
			break;
		case KFD_IOCTL_SVM_ATTR_ACCESS:
		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
			gpuidx = kfd_process_gpuidx_from_gpuid(p, val);
			break;
		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
			break;
		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
			break;
		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
			break;
		default:
			pr_debug("unknown attr type 0x%x\n", attrs[i].type);
			return -EINVAL;
		}

		if (gpuidx < 0) {
			pr_debug("no GPU 0x%x found\n", val);
			return -EINVAL;
		} else if (gpuidx < MAX_GPU_INSTANCE &&
			   !test_bit(gpuidx, p->svms.bitmap_supported)) {
			pr_debug("GPU 0x%x not supported\n", val);
			return -EINVAL;
		}
	}

	return 0;
}

static void
svm_range_apply_attrs(struct kfd_process *p, struct svm_range *prange,
		      uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs)
{
	uint32_t i;
	int gpuidx;

	for (i = 0; i < nattr; i++) {
		switch (attrs[i].type) {
		case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC:
			prange->preferred_loc = attrs[i].value;
			break;
		case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC:
			prange->prefetch_loc = attrs[i].value;
			break;
		case KFD_IOCTL_SVM_ATTR_ACCESS:
		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
			gpuidx = kfd_process_gpuidx_from_gpuid(p,
							       attrs[i].value);
			if (attrs[i].type == KFD_IOCTL_SVM_ATTR_NO_ACCESS) {
				bitmap_clear(prange->bitmap_access, gpuidx, 1);
				bitmap_clear(prange->bitmap_aip, gpuidx, 1);
			} else if (attrs[i].type == KFD_IOCTL_SVM_ATTR_ACCESS) {
				bitmap_set(prange->bitmap_access, gpuidx, 1);
				bitmap_clear(prange->bitmap_aip, gpuidx, 1);
			} else {
				bitmap_clear(prange->bitmap_access, gpuidx, 1);
				bitmap_set(prange->bitmap_aip, gpuidx, 1);
			}
			break;
		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
			prange->flags |= attrs[i].value;
			break;
		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
			prange->flags &= ~attrs[i].value;
			break;
		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
			prange->granularity = attrs[i].value;
			break;
		default:
			WARN_ONCE(1, "svm_range_check_attrs wasn't called?");
		}
	}
}

static bool
svm_range_is_same_attrs(struct kfd_process *p, struct svm_range *prange,
			uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs)
{
	uint32_t i;
	int gpuidx;

	for (i = 0; i < nattr; i++) {
		switch (attrs[i].type) {
		case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC:
			if (prange->preferred_loc != attrs[i].value)
				return false;
			break;
		case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC:
			/* Prefetch should always trigger a migration even
			 * if the value of the attribute didn't change.
			 */
			return false;
		case KFD_IOCTL_SVM_ATTR_ACCESS:
		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
			gpuidx = kfd_process_gpuidx_from_gpuid(p,
							       attrs[i].value);
			if (attrs[i].type == KFD_IOCTL_SVM_ATTR_NO_ACCESS) {
				if (test_bit(gpuidx, prange->bitmap_access) ||
				    test_bit(gpuidx, prange->bitmap_aip))
					return false;
			} else if (attrs[i].type == KFD_IOCTL_SVM_ATTR_ACCESS) {
				if (!test_bit(gpuidx, prange->bitmap_access))
					return false;
			} else {
				if (!test_bit(gpuidx, prange->bitmap_aip))
					return false;
			}
			break;
		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
			if ((prange->flags & attrs[i].value) != attrs[i].value)
				return false;
			break;
		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
			if ((prange->flags & attrs[i].value) != 0)
				return false;
			break;
		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
			if (prange->granularity != attrs[i].value)
				return false;
			break;
		default:
			WARN_ONCE(1, "svm_range_check_attrs wasn't called?");
		}
	}

	return true;
}

/**
 * svm_range_debug_dump - print all range information from svms
 * @svms: svm range list header
 *
 * debug output svm range start, end, prefetch location from svms
 * interval tree and link list
 *
 * Context: The caller must hold svms->lock
 */
static void svm_range_debug_dump(struct svm_range_list *svms)
{
	struct interval_tree_node *node;
	struct svm_range *prange;

	pr_debug("dump svms 0x%p list\n", svms);
	pr_debug("range\tstart\tpage\tend\t\tlocation\n");

	list_for_each_entry(prange, &svms->list, list) {
		pr_debug("0x%p 0x%lx\t0x%llx\t0x%llx\t0x%x\n",
			 prange, prange->start, prange->npages,
			 prange->start + prange->npages - 1,
			 prange->actual_loc);
	}

	pr_debug("dump svms 0x%p interval tree\n", svms);
	pr_debug("range\tstart\tpage\tend\t\tlocation\n");
	node = interval_tree_iter_first(&svms->objects, 0, ~0ULL);
	while (node) {
		prange = container_of(node, struct svm_range, it_node);
		pr_debug("0x%p 0x%lx\t0x%llx\t0x%llx\t0x%x\n",
			 prange, prange->start, prange->npages,
			 prange->start + prange->npages - 1,
			 prange->actual_loc);
		node = interval_tree_iter_next(node, 0, ~0ULL);
	}
}

static int
svm_range_split_array(void *ppnew, void *ppold, size_t size,
		      uint64_t old_start, uint64_t old_n,
		      uint64_t new_start, uint64_t new_n)
{
	unsigned char *new, *old, *pold;
	uint64_t d;

	if (!ppold)
		return 0;
	pold = *(unsigned char **)ppold;
	if (!pold)
		return 0;

	new = kvmalloc_array(new_n, size, GFP_KERNEL);
	if (!new)
		return -ENOMEM;

	d = (new_start - old_start) * size;
	memcpy(new, pold + d, new_n * size);

	old = kvmalloc_array(old_n, size, GFP_KERNEL);
	if (!old) {
		kvfree(new);
		return -ENOMEM;
	}

	d = (new_start == old_start) ? new_n * size : 0;
	memcpy(old, pold + d, old_n * size);

	kvfree(pold);
	*(void **)ppold = old;
	*(void **)ppnew = new;

	return 0;
}

static int
svm_range_split_pages(struct svm_range *new, struct svm_range *old,
		      uint64_t start, uint64_t last)
{
	uint64_t npages = last - start + 1;
	int i, r;

	for (i = 0; i < MAX_GPU_INSTANCE; i++) {
		r = svm_range_split_array(&new->dma_addr[i], &old->dma_addr[i],
					  sizeof(*old->dma_addr[i]), old->start,
					  npages, new->start, new->npages);
		if (r)
			return r;
	}

	return 0;
}

static int
svm_range_split_nodes(struct svm_range *new, struct svm_range *old,
		      uint64_t start, uint64_t last)
{
	uint64_t npages = last - start + 1;

	pr_debug("svms 0x%p new prange 0x%p start 0x%lx [0x%llx 0x%llx]\n",
		 new->svms, new, new->start, start, last);

	if (new->start == old->start) {
		new->offset = old->offset;
		old->offset += new->npages;
	} else {
		new->offset = old->offset + npages;
	}

	new->svm_bo = svm_range_bo_ref(old->svm_bo);
	new->ttm_res = old->ttm_res;

	spin_lock(&new->svm_bo->list_lock);
	list_add(&new->svm_bo_list, &new->svm_bo->range_list);
	spin_unlock(&new->svm_bo->list_lock);

	return 0;
}

/**
 * svm_range_split_adjust - split range and adjust
 *
 * @new: new range
 * @old: the old range
 * @start: the old range adjust to start address in pages
 * @last: the old range adjust to last address in pages
 *
 * Copy system memory dma_addr or vram ttm_res in old range to new
 * range from new_start up to size new->npages, the remaining old range is from
 * start to last
 *
 * Return:
 * 0 - OK, -ENOMEM - out of memory
 */
static int
svm_range_split_adjust(struct svm_range *new, struct svm_range *old,
		      uint64_t start, uint64_t last)
{
	int r;

	pr_debug("svms 0x%p new 0x%lx old [0x%lx 0x%lx] => [0x%llx 0x%llx]\n",
		 new->svms, new->start, old->start, old->last, start, last);

	if (new->start < old->start ||
	    new->last > old->last) {
		WARN_ONCE(1, "invalid new range start or last\n");
		return -EINVAL;
	}

	r = svm_range_split_pages(new, old, start, last);
	if (r)
		return r;

	if (old->actual_loc && old->ttm_res) {
		r = svm_range_split_nodes(new, old, start, last);
		if (r)
			return r;
	}

	old->npages = last - start + 1;
	old->start = start;
	old->last = last;
	new->flags = old->flags;
	new->preferred_loc = old->preferred_loc;
	new->prefetch_loc = old->prefetch_loc;
	new->actual_loc = old->actual_loc;
	new->granularity = old->granularity;
	bitmap_copy(new->bitmap_access, old->bitmap_access, MAX_GPU_INSTANCE);
	bitmap_copy(new->bitmap_aip, old->bitmap_aip, MAX_GPU_INSTANCE);

	return 0;
}

/**
 * svm_range_split - split a range in 2 ranges
 *
 * @prange: the svm range to split
 * @start: the remaining range start address in pages
 * @last: the remaining range last address in pages
 * @new: the result new range generated
 *
 * Two cases only:
 * case 1: if start == prange->start
 *         prange ==> prange[start, last]
 *         new range [last + 1, prange->last]
 *
 * case 2: if last == prange->last
 *         prange ==> prange[start, last]
 *         new range [prange->start, start - 1]
 *
 * Return:
 * 0 - OK, -ENOMEM - out of memory, -EINVAL - invalid start, last
 */
static int
svm_range_split(struct svm_range *prange, uint64_t start, uint64_t last,
		struct svm_range **new)
{
	uint64_t old_start = prange->start;
	uint64_t old_last = prange->last;
	struct svm_range_list *svms;
	int r = 0;

	pr_debug("svms 0x%p [0x%llx 0x%llx] to [0x%llx 0x%llx]\n", prange->svms,
		 old_start, old_last, start, last);

	if (old_start != start && old_last != last)
		return -EINVAL;
	if (start < old_start || last > old_last)
		return -EINVAL;

	svms = prange->svms;
	if (old_start == start)
		*new = svm_range_new(svms, last + 1, old_last);
	else
		*new = svm_range_new(svms, old_start, start - 1);
	if (!*new)
		return -ENOMEM;

	r = svm_range_split_adjust(*new, prange, start, last);
	if (r) {
		pr_debug("failed %d split [0x%llx 0x%llx] to [0x%llx 0x%llx]\n",
			 r, old_start, old_last, start, last);
		svm_range_free(*new);
		*new = NULL;
	}

	return r;
}

static int
svm_range_split_tail(struct svm_range *prange,
		     uint64_t new_last, struct list_head *insert_list)
{
	struct svm_range *tail;
	int r = svm_range_split(prange, prange->start, new_last, &tail);

	if (!r)
		list_add(&tail->list, insert_list);
	return r;
}

static int
svm_range_split_head(struct svm_range *prange,
		     uint64_t new_start, struct list_head *insert_list)
{
	struct svm_range *head;
	int r = svm_range_split(prange, new_start, prange->last, &head);

	if (!r)
		list_add(&head->list, insert_list);
	return r;
}

static void
svm_range_add_child(struct svm_range *prange, struct mm_struct *mm,
		    struct svm_range *pchild, enum svm_work_list_ops op)
{
	pr_debug("add child 0x%p [0x%lx 0x%lx] to prange 0x%p child list %d\n",
		 pchild, pchild->start, pchild->last, prange, op);

	pchild->work_item.mm = mm;
	pchild->work_item.op = op;
	list_add_tail(&pchild->child_list, &prange->child_list);
}

/**
 * svm_range_split_by_granularity - collect ranges within granularity boundary
 *
 * @p: the process with svms list
 * @mm: mm structure
 * @addr: the vm fault address in pages, to split the prange
 * @parent: parent range if prange is from child list
 * @prange: prange to split
 *
 * Trims @prange to be a single aligned block of prange->granularity if
 * possible. The head and tail are added to the child_list in @parent.
 *
 * Context: caller must hold mmap_read_lock and prange->lock
 *
 * Return:
 * 0 - OK, otherwise error code
 */
int
svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm,
			       unsigned long addr, struct svm_range *parent,
			       struct svm_range *prange)
{
	struct svm_range *head, *tail;
	unsigned long start, last, size;
	int r;

	/* Align splited range start and size to granularity size, then a single
	 * PTE will be used for whole range, this reduces the number of PTE
	 * updated and the L1 TLB space used for translation.
	 */
	size = 1UL << prange->granularity;
	start = ALIGN_DOWN(addr, size);
	last = ALIGN(addr + 1, size) - 1;

	pr_debug("svms 0x%p split [0x%lx 0x%lx] to [0x%lx 0x%lx] size 0x%lx\n",
		 prange->svms, prange->start, prange->last, start, last, size);

	if (start > prange->start) {
		r = svm_range_split(prange, start, prange->last, &head);
		if (r)
			return r;
		svm_range_add_child(parent, mm, head, SVM_OP_ADD_RANGE);
	}

	if (last < prange->last) {
		r = svm_range_split(prange, prange->start, last, &tail);
		if (r)
			return r;
		svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE);
	}

	/* xnack on, update mapping on GPUs with ACCESS_IN_PLACE */
	if (p->xnack_enabled && prange->work_item.op == SVM_OP_ADD_RANGE) {
		prange->work_item.op = SVM_OP_ADD_RANGE_AND_MAP;
		pr_debug("change prange 0x%p [0x%lx 0x%lx] op %d\n",
			 prange, prange->start, prange->last,
			 SVM_OP_ADD_RANGE_AND_MAP);
	}
	return 0;
}

static uint64_t
svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
			int domain)
{
	struct amdgpu_device *bo_adev;
	uint32_t flags = prange->flags;
	uint32_t mapping_flags = 0;
	uint64_t pte_flags;
	bool snoop = (domain != SVM_RANGE_VRAM_DOMAIN);
	bool coherent = flags & KFD_IOCTL_SVM_FLAG_COHERENT;

	if (domain == SVM_RANGE_VRAM_DOMAIN)
		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);

	switch (KFD_GC_VERSION(adev->kfd.dev)) {
	case IP_VERSION(9, 4, 1):
		if (domain == SVM_RANGE_VRAM_DOMAIN) {
			if (bo_adev == adev) {
				mapping_flags |= coherent ?
					AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
			} else {
				mapping_flags |= coherent ?
					AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
				if (amdgpu_xgmi_same_hive(adev, bo_adev))
					snoop = true;
			}
		} else {
			mapping_flags |= coherent ?
				AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
		}
		break;
	case IP_VERSION(9, 4, 2):
		if (domain == SVM_RANGE_VRAM_DOMAIN) {
			if (bo_adev == adev) {
				mapping_flags |= coherent ?
					AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
				if (adev->gmc.xgmi.connected_to_cpu)
					snoop = true;
			} else {
				mapping_flags |= coherent ?
					AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
				if (amdgpu_xgmi_same_hive(adev, bo_adev))
					snoop = true;
			}
		} else {
			mapping_flags |= coherent ?
				AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
		}
		break;
	default:
		mapping_flags |= coherent ?
			AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
	}

	mapping_flags |= AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE;

	if (flags & KFD_IOCTL_SVM_FLAG_GPU_RO)
		mapping_flags &= ~AMDGPU_VM_PAGE_WRITEABLE;
	if (flags & KFD_IOCTL_SVM_FLAG_GPU_EXEC)
		mapping_flags |= AMDGPU_VM_PAGE_EXECUTABLE;

	pte_flags = AMDGPU_PTE_VALID;
	pte_flags |= (domain == SVM_RANGE_VRAM_DOMAIN) ? 0 : AMDGPU_PTE_SYSTEM;
	pte_flags |= snoop ? AMDGPU_PTE_SNOOPED : 0;

	pte_flags |= amdgpu_gem_va_map_flags(adev, mapping_flags);
	return pte_flags;
}

static int
svm_range_unmap_from_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
			 uint64_t start, uint64_t last,
			 struct dma_fence **fence)
{
	uint64_t init_pte_value = 0;

	pr_debug("[0x%llx 0x%llx]\n", start, last);

	return amdgpu_vm_bo_update_mapping(adev, adev, vm, false, true, NULL,
					   start, last, init_pte_value, 0,
					   NULL, NULL, fence, NULL);
}

static int
svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start,
			  unsigned long last)
{
	DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE);
	struct kfd_process_device *pdd;
	struct dma_fence *fence = NULL;
	struct kfd_process *p;
	uint32_t gpuidx;
	int r = 0;

	bitmap_or(bitmap, prange->bitmap_access, prange->bitmap_aip,
		  MAX_GPU_INSTANCE);
	p = container_of(prange->svms, struct kfd_process, svms);

	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
		pr_debug("unmap from gpu idx 0x%x\n", gpuidx);
		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
		if (!pdd) {
			pr_debug("failed to find device idx %d\n", gpuidx);
			return -EINVAL;
		}

		r = svm_range_unmap_from_gpu(pdd->dev->adev,
					     drm_priv_to_vm(pdd->drm_priv),
					     start, last, &fence);
		if (r)
			break;

		if (fence) {
			r = dma_fence_wait(fence, false);
			dma_fence_put(fence);
			fence = NULL;
			if (r)
				break;
		}
		amdgpu_amdkfd_flush_gpu_tlb_pasid(pdd->dev->adev,
					p->pasid, TLB_FLUSH_HEAVYWEIGHT);
	}

	return r;
}

static int
svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
		     struct svm_range *prange, unsigned long offset,
		     unsigned long npages, bool readonly, dma_addr_t *dma_addr,
		     struct amdgpu_device *bo_adev, struct dma_fence **fence)
{
	bool table_freed = false;
	uint64_t pte_flags;
	unsigned long last_start;
	int last_domain;
	int r = 0;
	int64_t i, j;

	last_start = prange->start + offset;

	pr_debug("svms 0x%p [0x%lx 0x%lx] readonly %d\n", prange->svms,
		 last_start, last_start + npages - 1, readonly);

	for (i = offset; i < offset + npages; i++) {
		last_domain = dma_addr[i] & SVM_RANGE_VRAM_DOMAIN;
		dma_addr[i] &= ~SVM_RANGE_VRAM_DOMAIN;

		/* Collect all pages in the same address range and memory domain
		 * that can be mapped with a single call to update mapping.
		 */
		if (i < offset + npages - 1 &&
		    last_domain == (dma_addr[i + 1] & SVM_RANGE_VRAM_DOMAIN))
			continue;

		pr_debug("Mapping range [0x%lx 0x%llx] on domain: %s\n",
			 last_start, prange->start + i, last_domain ? "GPU" : "CPU");

		pte_flags = svm_range_get_pte_flags(adev, prange, last_domain);
		if (readonly)
			pte_flags &= ~AMDGPU_PTE_WRITEABLE;

		pr_debug("svms 0x%p map [0x%lx 0x%llx] vram %d PTE 0x%llx\n",
			 prange->svms, last_start, prange->start + i,
			 (last_domain == SVM_RANGE_VRAM_DOMAIN) ? 1 : 0,
			 pte_flags);

		r = amdgpu_vm_bo_update_mapping(adev, bo_adev, vm, false, false,
						NULL, last_start,
						prange->start + i, pte_flags,
						last_start - prange->start,
						NULL, dma_addr,
						&vm->last_update,
						&table_freed);

		for (j = last_start - prange->start; j <= i; j++)
			dma_addr[j] |= last_domain;

		if (r) {
			pr_debug("failed %d to map to gpu 0x%lx\n", r, prange->start);
			goto out;
		}
		last_start = prange->start + i + 1;
	}

	r = amdgpu_vm_update_pdes(adev, vm, false);
	if (r) {
		pr_debug("failed %d to update directories 0x%lx\n", r,
			 prange->start);
		goto out;
	}

	if (fence)
		*fence = dma_fence_get(vm->last_update);

	if (table_freed) {
		struct kfd_process *p;

		p = container_of(prange->svms, struct kfd_process, svms);
		amdgpu_amdkfd_flush_gpu_tlb_pasid(adev, p->pasid, TLB_FLUSH_LEGACY);
	}
out:
	return r;
}

static int
svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
		      unsigned long npages, bool readonly,
		      unsigned long *bitmap, bool wait)
{
	struct kfd_process_device *pdd;
	struct amdgpu_device *bo_adev;
	struct kfd_process *p;
	struct dma_fence *fence = NULL;
	uint32_t gpuidx;
	int r = 0;

	if (prange->svm_bo && prange->ttm_res)
		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
	else
		bo_adev = NULL;

	p = container_of(prange->svms, struct kfd_process, svms);
	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
		pr_debug("mapping to gpu idx 0x%x\n", gpuidx);
		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
		if (!pdd) {
			pr_debug("failed to find device idx %d\n", gpuidx);
			return -EINVAL;
		}

		pdd = kfd_bind_process_to_device(pdd->dev, p);
		if (IS_ERR(pdd))
			return -EINVAL;

		if (bo_adev && pdd->dev->adev != bo_adev &&
		    !amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) {
			pr_debug("cannot map to device idx %d\n", gpuidx);
			continue;
		}

		r = svm_range_map_to_gpu(pdd->dev->adev, drm_priv_to_vm(pdd->drm_priv),
					 prange, offset, npages, readonly,
					 prange->dma_addr[gpuidx],
					 bo_adev, wait ? &fence : NULL);
		if (r)
			break;

		if (fence) {
			r = dma_fence_wait(fence, false);
			dma_fence_put(fence);
			fence = NULL;
			if (r) {
				pr_debug("failed %d to dma fence wait\n", r);
				break;
			}
		}
	}

	return r;
}

struct svm_validate_context {
	struct kfd_process *process;
	struct svm_range *prange;
	bool intr;
	unsigned long bitmap[MAX_GPU_INSTANCE];
	struct ttm_validate_buffer tv[MAX_GPU_INSTANCE];
	struct list_head validate_list;
	struct ww_acquire_ctx ticket;
};

static int svm_range_reserve_bos(struct svm_validate_context *ctx)
{
	struct kfd_process_device *pdd;
	struct amdgpu_vm *vm;
	uint32_t gpuidx;
	int r;

	INIT_LIST_HEAD(&ctx->validate_list);
	for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) {
		pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx);
		if (!pdd) {
			pr_debug("failed to find device idx %d\n", gpuidx);
			return -EINVAL;
		}
		vm = drm_priv_to_vm(pdd->drm_priv);

		ctx->tv[gpuidx].bo = &vm->root.bo->tbo;
		ctx->tv[gpuidx].num_shared = 4;
		list_add(&ctx->tv[gpuidx].head, &ctx->validate_list);
	}

	r = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->validate_list,
				   ctx->intr, NULL);
	if (r) {
		pr_debug("failed %d to reserve bo\n", r);
		return r;
	}

	for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) {
		pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx);
		if (!pdd) {
			pr_debug("failed to find device idx %d\n", gpuidx);
			r = -EINVAL;
			goto unreserve_out;
		}

		r = amdgpu_vm_validate_pt_bos(pdd->dev->adev,
					      drm_priv_to_vm(pdd->drm_priv),
					      svm_range_bo_validate, NULL);
		if (r) {
			pr_debug("failed %d validate pt bos\n", r);
			goto unreserve_out;
		}
	}

	return 0;

unreserve_out:
	ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list);
	return r;
}

static void svm_range_unreserve_bos(struct svm_validate_context *ctx)
{
	ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list);
}

static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx)
{
	struct kfd_process_device *pdd;

	pdd = kfd_process_device_from_gpuidx(p, gpuidx);

	return SVM_ADEV_PGMAP_OWNER(pdd->dev->adev);
}

/*
 * Validation+GPU mapping with concurrent invalidation (MMU notifiers)
 *
 * To prevent concurrent destruction or change of range attributes, the
 * svm_read_lock must be held. The caller must not hold the svm_write_lock
 * because that would block concurrent evictions and lead to deadlocks. To
 * serialize concurrent migrations or validations of the same range, the
 * prange->migrate_mutex must be held.
 *
 * For VRAM ranges, the SVM BO must be allocated and valid (protected by its
 * eviction fence.
 *
 * The following sequence ensures race-free validation and GPU mapping:
 *
 * 1. Reserve page table (and SVM BO if range is in VRAM)
 * 2. hmm_range_fault to get page addresses (if system memory)
 * 3. DMA-map pages (if system memory)
 * 4-a. Take notifier lock
 * 4-b. Check that pages still valid (mmu_interval_read_retry)
 * 4-c. Check that the range was not split or otherwise invalidated
 * 4-d. Update GPU page table
 * 4.e. Release notifier lock
 * 5. Release page table (and SVM BO) reservation
 */
static int svm_range_validate_and_map(struct mm_struct *mm,
				      struct svm_range *prange,
				      int32_t gpuidx, bool intr, bool wait)
{
	struct svm_validate_context ctx;
	unsigned long start, end, addr;
	struct kfd_process *p;
	void *owner;
	int32_t idx;
	int r = 0;

	ctx.process = container_of(prange->svms, struct kfd_process, svms);
	ctx.prange = prange;
	ctx.intr = intr;

	if (gpuidx < MAX_GPU_INSTANCE) {
		bitmap_zero(ctx.bitmap, MAX_GPU_INSTANCE);
		bitmap_set(ctx.bitmap, gpuidx, 1);
	} else if (ctx.process->xnack_enabled) {
		bitmap_copy(ctx.bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);

		/* If prefetch range to GPU, or GPU retry fault migrate range to
		 * GPU, which has ACCESS attribute to the range, create mapping
		 * on that GPU.
		 */
		if (prange->actual_loc) {
			gpuidx = kfd_process_gpuidx_from_gpuid(ctx.process,
							prange->actual_loc);
			if (gpuidx < 0) {
				WARN_ONCE(1, "failed get device by id 0x%x\n",
					 prange->actual_loc);
				return -EINVAL;
			}
			if (test_bit(gpuidx, prange->bitmap_access))
				bitmap_set(ctx.bitmap, gpuidx, 1);
		}
	} else {
		bitmap_or(ctx.bitmap, prange->bitmap_access,
			  prange->bitmap_aip, MAX_GPU_INSTANCE);
	}

	if (bitmap_empty(ctx.bitmap, MAX_GPU_INSTANCE))
		return 0;

	if (prange->actual_loc && !prange->ttm_res) {
		/* This should never happen. actual_loc gets set by
		 * svm_migrate_ram_to_vram after allocating a BO.
		 */
		WARN_ONCE(1, "VRAM BO missing during validation\n");
		return -EINVAL;
	}

	svm_range_reserve_bos(&ctx);

	p = container_of(prange->svms, struct kfd_process, svms);
	owner = kfd_svm_page_owner(p, find_first_bit(ctx.bitmap,
						MAX_GPU_INSTANCE));
	for_each_set_bit(idx, ctx.bitmap, MAX_GPU_INSTANCE) {
		if (kfd_svm_page_owner(p, idx) != owner) {
			owner = NULL;
			break;
		}
	}

	start = prange->start << PAGE_SHIFT;
	end = (prange->last + 1) << PAGE_SHIFT;
	for (addr = start; addr < end && !r; ) {
		struct hmm_range *hmm_range;
		struct vm_area_struct *vma;
		unsigned long next;
		unsigned long offset;
		unsigned long npages;
		bool readonly;

		vma = find_vma(mm, addr);
		if (!vma || addr < vma->vm_start) {
			r = -EFAULT;
			goto unreserve_out;
		}
		readonly = !(vma->vm_flags & VM_WRITE);

		next = min(vma->vm_end, end);
		npages = (next - addr) >> PAGE_SHIFT;
		WRITE_ONCE(p->svms.faulting_task, current);
		r = amdgpu_hmm_range_get_pages(&prange->notifier, mm, NULL,
					       addr, npages, &hmm_range,
					       readonly, true, owner);
		WRITE_ONCE(p->svms.faulting_task, NULL);
		if (r) {
			pr_debug("failed %d to get svm range pages\n", r);
			goto unreserve_out;
		}

		offset = (addr - start) >> PAGE_SHIFT;
		r = svm_range_dma_map(prange, ctx.bitmap, offset, npages,
				      hmm_range->hmm_pfns);
		if (r) {
			pr_debug("failed %d to dma map range\n", r);
			goto unreserve_out;
		}

		svm_range_lock(prange);
		if (amdgpu_hmm_range_get_pages_done(hmm_range)) {
			pr_debug("hmm update the range, need validate again\n");
			r = -EAGAIN;
			goto unlock_out;
		}
		if (!list_empty(&prange->child_list)) {
			pr_debug("range split by unmap in parallel, validate again\n");
			r = -EAGAIN;
			goto unlock_out;
		}

		r = svm_range_map_to_gpus(prange, offset, npages, readonly,
					  ctx.bitmap, wait);

unlock_out:
		svm_range_unlock(prange);

		addr = next;
	}

	if (addr == end)
		prange->validated_once = true;

unreserve_out:
	svm_range_unreserve_bos(&ctx);

	if (!r)
		prange->validate_timestamp = ktime_to_us(ktime_get());

	return r;
}

/**
 * svm_range_list_lock_and_flush_work - flush pending deferred work
 *
 * @svms: the svm range list
 * @mm: the mm structure
 *
 * Context: Returns with mmap write lock held, pending deferred work flushed
 *
 */
void
svm_range_list_lock_and_flush_work(struct svm_range_list *svms,
				   struct mm_struct *mm)
{
retry_flush_work:
	flush_work(&svms->deferred_list_work);
	mmap_write_lock(mm);

	if (list_empty(&svms->deferred_range_list))
		return;
	mmap_write_unlock(mm);
	pr_debug("retry flush\n");
	goto retry_flush_work;
}

static void svm_range_restore_work(struct work_struct *work)
{
	struct delayed_work *dwork = to_delayed_work(work);
	struct svm_range_list *svms;
	struct svm_range *prange;
	struct kfd_process *p;
	struct mm_struct *mm;
	int evicted_ranges;
	int invalid;
	int r;

	svms = container_of(dwork, struct svm_range_list, restore_work);
	evicted_ranges = atomic_read(&svms->evicted_ranges);
	if (!evicted_ranges)
		return;

	pr_debug("restore svm ranges\n");

	/* kfd_process_notifier_release destroys this worker thread. So during
	 * the lifetime of this thread, kfd_process and mm will be valid.
	 */
	p = container_of(svms, struct kfd_process, svms);
	mm = p->mm;
	if (!mm)
		return;

	svm_range_list_lock_and_flush_work(svms, mm);
	mutex_lock(&svms->lock);

	evicted_ranges = atomic_read(&svms->evicted_ranges);

	list_for_each_entry(prange, &svms->list, list) {
		invalid = atomic_read(&prange->invalid);
		if (!invalid)
			continue;

		pr_debug("restoring svms 0x%p prange 0x%p [0x%lx %lx] inv %d\n",
			 prange->svms, prange, prange->start, prange->last,
			 invalid);

		/*
		 * If range is migrating, wait for migration is done.
		 */
		mutex_lock(&prange->migrate_mutex);

		r = svm_range_validate_and_map(mm, prange, MAX_GPU_INSTANCE,
					       false, true);
		if (r)
			pr_debug("failed %d to map 0x%lx to gpus\n", r,
				 prange->start);

		mutex_unlock(&prange->migrate_mutex);
		if (r)
			goto out_reschedule;

		if (atomic_cmpxchg(&prange->invalid, invalid, 0) != invalid)
			goto out_reschedule;
	}

	if (atomic_cmpxchg(&svms->evicted_ranges, evicted_ranges, 0) !=
	    evicted_ranges)
		goto out_reschedule;

	evicted_ranges = 0;

	r = kgd2kfd_resume_mm(mm);
	if (r) {
		/* No recovery from this failure. Probably the CP is
		 * hanging. No point trying again.
		 */
		pr_debug("failed %d to resume KFD\n", r);
	}

	pr_debug("restore svm ranges successfully\n");

out_reschedule:
	mutex_unlock(&svms->lock);
	mmap_write_unlock(mm);

	/* If validation failed, reschedule another attempt */
	if (evicted_ranges) {
		pr_debug("reschedule to restore svm range\n");
		schedule_delayed_work(&svms->restore_work,
			msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS));
	}
}

/**
 * svm_range_evict - evict svm range
 * @prange: svm range structure
 * @mm: current process mm_struct
 * @start: starting process queue number
 * @last: last process queue number
 *
 * Stop all queues of the process to ensure GPU doesn't access the memory, then
 * return to let CPU evict the buffer and proceed CPU pagetable update.
 *
 * Don't need use lock to sync cpu pagetable invalidation with GPU execution.
 * If invalidation happens while restore work is running, restore work will
 * restart to ensure to get the latest CPU pages mapping to GPU, then start
 * the queues.
 */
static int
svm_range_evict(struct svm_range *prange, struct mm_struct *mm,
		unsigned long start, unsigned long last)
{
	struct svm_range_list *svms = prange->svms;
	struct svm_range *pchild;
	struct kfd_process *p;
	int r = 0;

	p = container_of(svms, struct kfd_process, svms);

	pr_debug("invalidate svms 0x%p prange [0x%lx 0x%lx] [0x%lx 0x%lx]\n",
		 svms, prange->start, prange->last, start, last);

	if (!p->xnack_enabled) {
		int evicted_ranges;

		list_for_each_entry(pchild, &prange->child_list, child_list) {
			mutex_lock_nested(&pchild->lock, 1);
			if (pchild->start <= last && pchild->last >= start) {
				pr_debug("increment pchild invalid [0x%lx 0x%lx]\n",
					 pchild->start, pchild->last);
				atomic_inc(&pchild->invalid);
			}
			mutex_unlock(&pchild->lock);
		}

		if (prange->start <= last && prange->last >= start)
			atomic_inc(&prange->invalid);

		evicted_ranges = atomic_inc_return(&svms->evicted_ranges);
		if (evicted_ranges != 1)
			return r;

		pr_debug("evicting svms 0x%p range [0x%lx 0x%lx]\n",
			 prange->svms, prange->start, prange->last);

		/* First eviction, stop the queues */
		r = kgd2kfd_quiesce_mm(mm);
		if (r)
			pr_debug("failed to quiesce KFD\n");

		pr_debug("schedule to restore svm %p ranges\n", svms);
		schedule_delayed_work(&svms->restore_work,
			msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS));
	} else {
		unsigned long s, l;

		pr_debug("invalidate unmap svms 0x%p [0x%lx 0x%lx] from GPUs\n",
			 prange->svms, start, last);
		list_for_each_entry(pchild, &prange->child_list, child_list) {
			mutex_lock_nested(&pchild->lock, 1);
			s = max(start, pchild->start);
			l = min(last, pchild->last);
			if (l >= s)
				svm_range_unmap_from_gpus(pchild, s, l);
			mutex_unlock(&pchild->lock);
		}
		s = max(start, prange->start);
		l = min(last, prange->last);
		if (l >= s)
			svm_range_unmap_from_gpus(prange, s, l);
	}

	return r;
}

static struct svm_range *svm_range_clone(struct svm_range *old)
{
	struct svm_range *new;

	new = svm_range_new(old->svms, old->start, old->last);
	if (!new)
		return NULL;

	if (old->svm_bo) {
		new->ttm_res = old->ttm_res;
		new->offset = old->offset;
		new->svm_bo = svm_range_bo_ref(old->svm_bo);
		spin_lock(&new->svm_bo->list_lock);
		list_add(&new->svm_bo_list, &new->svm_bo->range_list);
		spin_unlock(&new->svm_bo->list_lock);
	}
	new->flags = old->flags;
	new->preferred_loc = old->preferred_loc;
	new->prefetch_loc = old->prefetch_loc;
	new->actual_loc = old->actual_loc;
	new->granularity = old->granularity;
	bitmap_copy(new->bitmap_access, old->bitmap_access, MAX_GPU_INSTANCE);
	bitmap_copy(new->bitmap_aip, old->bitmap_aip, MAX_GPU_INSTANCE);

	return new;
}

/**
 * svm_range_add - add svm range and handle overlap
 * @p: the range add to this process svms
 * @start: page size aligned
 * @size: page size aligned
 * @nattr: number of attributes
 * @attrs: array of attributes
 * @update_list: output, the ranges need validate and update GPU mapping
 * @insert_list: output, the ranges need insert to svms
 * @remove_list: output, the ranges are replaced and need remove from svms
 *
 * Check if the virtual address range has overlap with any existing ranges,
 * split partly overlapping ranges and add new ranges in the gaps. All changes
 * should be applied to the range_list and interval tree transactionally. If
 * any range split or allocation fails, the entire update fails. Therefore any
 * existing overlapping svm_ranges are cloned and the original svm_ranges left
 * unchanged.
 *
 * If the transaction succeeds, the caller can update and insert clones and
 * new ranges, then free the originals.
 *
 * Otherwise the caller can free the clones and new ranges, while the old
 * svm_ranges remain unchanged.
 *
 * Context: Process context, caller must hold svms->lock
 *
 * Return:
 * 0 - OK, otherwise error code
 */
static int
svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size,
	      uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs,
	      struct list_head *update_list, struct list_head *insert_list,
	      struct list_head *remove_list)
{
	unsigned long last = start + size - 1UL;
	struct svm_range_list *svms = &p->svms;
	struct interval_tree_node *node;
	struct svm_range *prange;
	struct svm_range *tmp;
	int r = 0;

	pr_debug("svms 0x%p [0x%llx 0x%lx]\n", &p->svms, start, last);

	INIT_LIST_HEAD(update_list);
	INIT_LIST_HEAD(insert_list);
	INIT_LIST_HEAD(remove_list);

	node = interval_tree_iter_first(&svms->objects, start, last);
	while (node) {
		struct interval_tree_node *next;
		unsigned long next_start;

		pr_debug("found overlap node [0x%lx 0x%lx]\n", node->start,
			 node->last);

		prange = container_of(node, struct svm_range, it_node);
		next = interval_tree_iter_next(node, start, last);
		next_start = min(node->last, last) + 1;

		if (svm_range_is_same_attrs(p, prange, nattr, attrs)) {
			/* nothing to do */
		} else if (node->start < start || node->last > last) {
			/* node intersects the update range and its attributes
			 * will change. Clone and split it, apply updates only
			 * to the overlapping part
			 */
			struct svm_range *old = prange;

			prange = svm_range_clone(old);
			if (!prange) {
				r = -ENOMEM;
				goto out;
			}

			list_add(&old->update_list, remove_list);
			list_add(&prange->list, insert_list);
			list_add(&prange->update_list, update_list);

			if (node->start < start) {
				pr_debug("change old range start\n");
				r = svm_range_split_head(prange, start,
							 insert_list);
				if (r)
					goto out;
			}
			if (node->last > last) {
				pr_debug("change old range last\n");
				r = svm_range_split_tail(prange, last,
							 insert_list);
				if (r)
					goto out;
			}
		} else {
			/* The node is contained within start..last,
			 * just update it
			 */
			list_add(&prange->update_list, update_list);
		}

		/* insert a new node if needed */
		if (node->start > start) {
			prange = svm_range_new(svms, start, node->start - 1);
			if (!prange) {
				r = -ENOMEM;
				goto out;
			}

			list_add(&prange->list, insert_list);
			list_add(&prange->update_list, update_list);
		}

		node = next;
		start = next_start;
	}

	/* add a final range at the end if needed */
	if (start <= last) {
		prange = svm_range_new(svms, start, last);
		if (!prange) {
			r = -ENOMEM;
			goto out;
		}
		list_add(&prange->list, insert_list);
		list_add(&prange->update_list, update_list);
	}

out:
	if (r)
		list_for_each_entry_safe(prange, tmp, insert_list, list)
			svm_range_free(prange);

	return r;
}

static void
svm_range_update_notifier_and_interval_tree(struct mm_struct *mm,
					    struct svm_range *prange)
{
	unsigned long start;
	unsigned long last;

	start = prange->notifier.interval_tree.start >> PAGE_SHIFT;
	last = prange->notifier.interval_tree.last >> PAGE_SHIFT;

	if (prange->start == start && prange->last == last)
		return;

	pr_debug("up notifier 0x%p prange 0x%p [0x%lx 0x%lx] [0x%lx 0x%lx]\n",
		  prange->svms, prange, start, last, prange->start,
		  prange->last);

	if (start != 0 && last != 0) {
		interval_tree_remove(&prange->it_node, &prange->svms->objects);
		svm_range_remove_notifier(prange);
	}
	prange->it_node.start = prange->start;
	prange->it_node.last = prange->last;

	interval_tree_insert(&prange->it_node, &prange->svms->objects);
	svm_range_add_notifier_locked(mm, prange);
}

static void
svm_range_handle_list_op(struct svm_range_list *svms, struct svm_range *prange)
{
	struct mm_struct *mm = prange->work_item.mm;

	switch (prange->work_item.op) {
	case SVM_OP_NULL:
		pr_debug("NULL OP 0x%p prange 0x%p [0x%lx 0x%lx]\n",
			 svms, prange, prange->start, prange->last);
		break;
	case SVM_OP_UNMAP_RANGE:
		pr_debug("remove 0x%p prange 0x%p [0x%lx 0x%lx]\n",
			 svms, prange, prange->start, prange->last);
		svm_range_unlink(prange);
		svm_range_remove_notifier(prange);
		svm_range_free(prange);
		break;
	case SVM_OP_UPDATE_RANGE_NOTIFIER:
		pr_debug("update notifier 0x%p prange 0x%p [0x%lx 0x%lx]\n",
			 svms, prange, prange->start, prange->last);
		svm_range_update_notifier_and_interval_tree(mm, prange);
		break;
	case SVM_OP_UPDATE_RANGE_NOTIFIER_AND_MAP:
		pr_debug("update and map 0x%p prange 0x%p [0x%lx 0x%lx]\n",
			 svms, prange, prange->start, prange->last);
		svm_range_update_notifier_and_interval_tree(mm, prange);
		/* TODO: implement deferred validation and mapping */
		break;
	case SVM_OP_ADD_RANGE:
		pr_debug("add 0x%p prange 0x%p [0x%lx 0x%lx]\n", svms, prange,
			 prange->start, prange->last);
		svm_range_add_to_svms(prange);
		svm_range_add_notifier_locked(mm, prange);
		break;
	case SVM_OP_ADD_RANGE_AND_MAP:
		pr_debug("add and map 0x%p prange 0x%p [0x%lx 0x%lx]\n", svms,
			 prange, prange->start, prange->last);
		svm_range_add_to_svms(prange);
		svm_range_add_notifier_locked(mm, prange);
		/* TODO: implement deferred validation and mapping */
		break;
	default:
		WARN_ONCE(1, "Unknown prange 0x%p work op %d\n", prange,
			 prange->work_item.op);
	}
}

static void svm_range_drain_retry_fault(struct svm_range_list *svms)
{
	struct kfd_process_device *pdd;
	struct kfd_process *p;
	int drain;
	uint32_t i;

	p = container_of(svms, struct kfd_process, svms);

restart:
	drain = atomic_read(&svms->drain_pagefaults);
	if (!drain)
		return;

	for_each_set_bit(i, svms->bitmap_supported, p->n_pdds) {
		pdd = p->pdds[i];
		if (!pdd)
			continue;

		pr_debug("drain retry fault gpu %d svms %p\n", i, svms);

		amdgpu_ih_wait_on_checkpoint_process_ts(pdd->dev->adev,
						     &pdd->dev->adev->irq.ih1);
		pr_debug("drain retry fault gpu %d svms 0x%p done\n", i, svms);
	}
	if (atomic_cmpxchg(&svms->drain_pagefaults, drain, 0) != drain)
		goto restart;
}

static void svm_range_deferred_list_work(struct work_struct *work)
{
	struct svm_range_list *svms;
	struct svm_range *prange;
	struct mm_struct *mm;
	struct kfd_process *p;

	svms = container_of(work, struct svm_range_list, deferred_list_work);
	pr_debug("enter svms 0x%p\n", svms);

	p = container_of(svms, struct kfd_process, svms);
	/* Avoid mm is gone when inserting mmu notifier */
	mm = get_task_mm(p->lead_thread);
	if (!mm) {
		pr_debug("svms 0x%p process mm gone\n", svms);
		return;
	}
retry:
	mmap_write_lock(mm);

	/* Checking for the need to drain retry faults must be inside
	 * mmap write lock to serialize with munmap notifiers.
	 */
	if (unlikely(atomic_read(&svms->drain_pagefaults))) {
		mmap_write_unlock(mm);
		svm_range_drain_retry_fault(svms);
		goto retry;
	}

	spin_lock(&svms->deferred_list_lock);
	while (!list_empty(&svms->deferred_range_list)) {
		prange = list_first_entry(&svms->deferred_range_list,
					  struct svm_range, deferred_list);
		list_del_init(&prange->deferred_list);
		spin_unlock(&svms->deferred_list_lock);

		pr_debug("prange 0x%p [0x%lx 0x%lx] op %d\n", prange,
			 prange->start, prange->last, prange->work_item.op);

		mutex_lock(&svms->lock);
		mutex_lock(&prange->migrate_mutex);
		while (!list_empty(&prange->child_list)) {
			struct svm_range *pchild;

			pchild = list_first_entry(&prange->child_list,
						struct svm_range, child_list);
			pr_debug("child prange 0x%p op %d\n", pchild,
				 pchild->work_item.op);
			list_del_init(&pchild->child_list);
			svm_range_handle_list_op(svms, pchild);
		}
		mutex_unlock(&prange->migrate_mutex);

		svm_range_handle_list_op(svms, prange);
		mutex_unlock(&svms->lock);

		spin_lock(&svms->deferred_list_lock);
	}
	spin_unlock(&svms->deferred_list_lock);

	mmap_write_unlock(mm);
	mmput(mm);
	pr_debug("exit svms 0x%p\n", svms);
}

void
svm_range_add_list_work(struct svm_range_list *svms, struct svm_range *prange,
			struct mm_struct *mm, enum svm_work_list_ops op)
{
	spin_lock(&svms->deferred_list_lock);
	/* if prange is on the deferred list */
	if (!list_empty(&prange->deferred_list)) {
		pr_debug("update exist prange 0x%p work op %d\n", prange, op);
		WARN_ONCE(prange->work_item.mm != mm, "unmatch mm\n");
		if (op != SVM_OP_NULL &&
		    prange->work_item.op != SVM_OP_UNMAP_RANGE)
			prange->work_item.op = op;
	} else {
		prange->work_item.op = op;
		prange->work_item.mm = mm;
		list_add_tail(&prange->deferred_list,
			      &prange->svms->deferred_range_list);
		pr_debug("add prange 0x%p [0x%lx 0x%lx] to work list op %d\n",
			 prange, prange->start, prange->last, op);
	}
	spin_unlock(&svms->deferred_list_lock);
}

void schedule_deferred_list_work(struct svm_range_list *svms)
{
	spin_lock(&svms->deferred_list_lock);
	if (!list_empty(&svms->deferred_range_list))
		schedule_work(&svms->deferred_list_work);
	spin_unlock(&svms->deferred_list_lock);
}

static void
svm_range_unmap_split(struct mm_struct *mm, struct svm_range *parent,
		      struct svm_range *prange, unsigned long start,
		      unsigned long last)
{
	struct svm_range *head;
	struct svm_range *tail;

	if (prange->work_item.op == SVM_OP_UNMAP_RANGE) {
		pr_debug("prange 0x%p [0x%lx 0x%lx] is already freed\n", prange,
			 prange->start, prange->last);
		return;
	}
	if (start > prange->last || last < prange->start)
		return;

	head = tail = prange;
	if (start > prange->start)
		svm_range_split(prange, prange->start, start - 1, &tail);
	if (last < tail->last)
		svm_range_split(tail, last + 1, tail->last, &head);

	if (head != prange && tail != prange) {
		svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE);
		svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE);
	} else if (tail != prange) {
		svm_range_add_child(parent, mm, tail, SVM_OP_UNMAP_RANGE);
	} else if (head != prange) {
		svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE);
	} else if (parent != prange) {
		prange->work_item.op = SVM_OP_UNMAP_RANGE;
	}
}

static void
svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange,
			 unsigned long start, unsigned long last)
{
	struct svm_range_list *svms;
	struct svm_range *pchild;
	struct kfd_process *p;
	unsigned long s, l;
	bool unmap_parent;

	p = kfd_lookup_process_by_mm(mm);
	if (!p)
		return;
	svms = &p->svms;

	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] [0x%lx 0x%lx]\n", svms,
		 prange, prange->start, prange->last, start, last);

	/* Make sure pending page faults are drained in the deferred worker
	 * before the range is freed to avoid straggler interrupts on
	 * unmapped memory causing "phantom faults".
	 */
	atomic_inc(&svms->drain_pagefaults);

	unmap_parent = start <= prange->start && last >= prange->last;

	list_for_each_entry(pchild, &prange->child_list, child_list) {
		mutex_lock_nested(&pchild->lock, 1);
		s = max(start, pchild->start);
		l = min(last, pchild->last);
		if (l >= s)
			svm_range_unmap_from_gpus(pchild, s, l);
		svm_range_unmap_split(mm, prange, pchild, start, last);
		mutex_unlock(&pchild->lock);
	}
	s = max(start, prange->start);
	l = min(last, prange->last);
	if (l >= s)
		svm_range_unmap_from_gpus(prange, s, l);
	svm_range_unmap_split(mm, prange, prange, start, last);

	if (unmap_parent)
		svm_range_add_list_work(svms, prange, mm, SVM_OP_UNMAP_RANGE);
	else
		svm_range_add_list_work(svms, prange, mm,
					SVM_OP_UPDATE_RANGE_NOTIFIER);
	schedule_deferred_list_work(svms);

	kfd_unref_process(p);
}

/**
 * svm_range_cpu_invalidate_pagetables - interval notifier callback
 * @mni: mmu_interval_notifier struct
 * @range: mmu_notifier_range struct
 * @cur_seq: value to pass to mmu_interval_set_seq()
 *
 * If event is MMU_NOTIFY_UNMAP, this is from CPU unmap range, otherwise, it
 * is from migration, or CPU page invalidation callback.
 *
 * For unmap event, unmap range from GPUs, remove prange from svms in a delayed
 * work thread, and split prange if only part of prange is unmapped.
 *
 * For invalidation event, if GPU retry fault is not enabled, evict the queues,
 * then schedule svm_range_restore_work to update GPU mapping and resume queues.
 * If GPU retry fault is enabled, unmap the svm range from GPU, retry fault will
 * update GPU mapping to recover.
 *
 * Context: mmap lock, notifier_invalidate_start lock are held
 *          for invalidate event, prange lock is held if this is from migration
 */
static bool
svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni,
				    const struct mmu_notifier_range *range,
				    unsigned long cur_seq)
{
	struct svm_range *prange;
	unsigned long start;
	unsigned long last;

	if (range->event == MMU_NOTIFY_RELEASE)
		return true;

	start = mni->interval_tree.start;
	last = mni->interval_tree.last;
	start = max(start, range->start) >> PAGE_SHIFT;
	last = min(last, range->end - 1) >> PAGE_SHIFT;
	pr_debug("[0x%lx 0x%lx] range[0x%lx 0x%lx] notifier[0x%lx 0x%lx] %d\n",
		 start, last, range->start >> PAGE_SHIFT,
		 (range->end - 1) >> PAGE_SHIFT,
		 mni->interval_tree.start >> PAGE_SHIFT,
		 mni->interval_tree.last >> PAGE_SHIFT, range->event);

	prange = container_of(mni, struct svm_range, notifier);

	svm_range_lock(prange);
	mmu_interval_set_seq(mni, cur_seq);

	switch (range->event) {
	case MMU_NOTIFY_UNMAP:
		svm_range_unmap_from_cpu(mni->mm, prange, start, last);
		break;
	default:
		svm_range_evict(prange, mni->mm, start, last);
		break;
	}

	svm_range_unlock(prange);

	return true;
}

/**
 * svm_range_from_addr - find svm range from fault address
 * @svms: svm range list header
 * @addr: address to search range interval tree, in pages
 * @parent: parent range if range is on child list
 *
 * Context: The caller must hold svms->lock
 *
 * Return: the svm_range found or NULL
 */
struct svm_range *
svm_range_from_addr(struct svm_range_list *svms, unsigned long addr,
		    struct svm_range **parent)
{
	struct interval_tree_node *node;
	struct svm_range *prange;
	struct svm_range *pchild;

	node = interval_tree_iter_first(&svms->objects, addr, addr);
	if (!node)
		return NULL;

	prange = container_of(node, struct svm_range, it_node);
	pr_debug("address 0x%lx prange [0x%lx 0x%lx] node [0x%lx 0x%lx]\n",
		 addr, prange->start, prange->last, node->start, node->last);

	if (addr >= prange->start && addr <= prange->last) {
		if (parent)
			*parent = prange;
		return prange;
	}
	list_for_each_entry(pchild, &prange->child_list, child_list)
		if (addr >= pchild->start && addr <= pchild->last) {
			pr_debug("found address 0x%lx pchild [0x%lx 0x%lx]\n",
				 addr, pchild->start, pchild->last);
			if (parent)
				*parent = prange;
			return pchild;
		}

	return NULL;
}

/* svm_range_best_restore_location - decide the best fault restore location
 * @prange: svm range structure
 * @adev: the GPU on which vm fault happened
 *
 * This is only called when xnack is on, to decide the best location to restore
 * the range mapping after GPU vm fault. Caller uses the best location to do
 * migration if actual loc is not best location, then update GPU page table
 * mapping to the best location.
 *
 * If the preferred loc is accessible by faulting GPU, use preferred loc.
 * If vm fault gpu idx is on range ACCESSIBLE bitmap, best_loc is vm fault gpu
 * If vm fault gpu idx is on range ACCESSIBLE_IN_PLACE bitmap, then
 *    if range actual loc is cpu, best_loc is cpu
 *    if vm fault gpu is on xgmi same hive of range actual loc gpu, best_loc is
 *    range actual loc.
 * Otherwise, GPU no access, best_loc is -1.
 *
 * Return:
 * -1 means vm fault GPU no access
 * 0 for CPU or GPU id
 */
static int32_t
svm_range_best_restore_location(struct svm_range *prange,
				struct amdgpu_device *adev,
				int32_t *gpuidx)
{
	struct amdgpu_device *bo_adev, *preferred_adev;
	struct kfd_process *p;
	uint32_t gpuid;
	int r;

	p = container_of(prange->svms, struct kfd_process, svms);

	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, gpuidx);
	if (r < 0) {
		pr_debug("failed to get gpuid from kgd\n");
		return -1;
	}

	if (prange->preferred_loc == gpuid ||
	    prange->preferred_loc == KFD_IOCTL_SVM_LOCATION_SYSMEM) {
		return prange->preferred_loc;
	} else if (prange->preferred_loc != KFD_IOCTL_SVM_LOCATION_UNDEFINED) {
		preferred_adev = svm_range_get_adev_by_id(prange,
							prange->preferred_loc);
		if (amdgpu_xgmi_same_hive(adev, preferred_adev))
			return prange->preferred_loc;
		/* fall through */
	}

	if (test_bit(*gpuidx, prange->bitmap_access))
		return gpuid;

	if (test_bit(*gpuidx, prange->bitmap_aip)) {
		if (!prange->actual_loc)
			return 0;

		bo_adev = svm_range_get_adev_by_id(prange, prange->actual_loc);
		if (amdgpu_xgmi_same_hive(adev, bo_adev))
			return prange->actual_loc;
		else
			return 0;
	}

	return -1;
}

static int
svm_range_get_range_boundaries(struct kfd_process *p, int64_t addr,
			       unsigned long *start, unsigned long *last,
			       bool *is_heap_stack)
{
	struct vm_area_struct *vma;
	struct interval_tree_node *node;
	unsigned long start_limit, end_limit;

	vma = find_vma(p->mm, addr << PAGE_SHIFT);
	if (!vma || (addr << PAGE_SHIFT) < vma->vm_start) {
		pr_debug("VMA does not exist in address [0x%llx]\n", addr);
		return -EFAULT;
	}

	*is_heap_stack = (vma->vm_start <= vma->vm_mm->brk &&
			  vma->vm_end >= vma->vm_mm->start_brk) ||
			 (vma->vm_start <= vma->vm_mm->start_stack &&
			  vma->vm_end >= vma->vm_mm->start_stack);

	start_limit = max(vma->vm_start >> PAGE_SHIFT,
		      (unsigned long)ALIGN_DOWN(addr, 2UL << 8));
	end_limit = min(vma->vm_end >> PAGE_SHIFT,
		    (unsigned long)ALIGN(addr + 1, 2UL << 8));
	/* First range that starts after the fault address */
	node = interval_tree_iter_first(&p->svms.objects, addr + 1, ULONG_MAX);
	if (node) {
		end_limit = min(end_limit, node->start);
		/* Last range that ends before the fault address */
		node = container_of(rb_prev(&node->rb),
				    struct interval_tree_node, rb);
	} else {
		/* Last range must end before addr because
		 * there was no range after addr
		 */
		node = container_of(rb_last(&p->svms.objects.rb_root),
				    struct interval_tree_node, rb);
	}
	if (node) {
		if (node->last >= addr) {
			WARN(1, "Overlap with prev node and page fault addr\n");
			return -EFAULT;
		}
		start_limit = max(start_limit, node->last + 1);
	}

	*start = start_limit;
	*last = end_limit - 1;

	pr_debug("vma [0x%lx 0x%lx] range [0x%lx 0x%lx] is_heap_stack %d\n",
		 vma->vm_start >> PAGE_SHIFT, vma->vm_end >> PAGE_SHIFT,
		 *start, *last, *is_heap_stack);

	return 0;
}

static int
svm_range_check_vm_userptr(struct kfd_process *p, uint64_t start, uint64_t last,
			   uint64_t *bo_s, uint64_t *bo_l)
{
	struct amdgpu_bo_va_mapping *mapping;
	struct interval_tree_node *node;
	struct amdgpu_bo *bo = NULL;
	unsigned long userptr;
	uint32_t i;
	int r;

	for (i = 0; i < p->n_pdds; i++) {
		struct amdgpu_vm *vm;

		if (!p->pdds[i]->drm_priv)
			continue;

		vm = drm_priv_to_vm(p->pdds[i]->drm_priv);
		r = amdgpu_bo_reserve(vm->root.bo, false);
		if (r)
			return r;

		/* Check userptr by searching entire vm->va interval tree */
		node = interval_tree_iter_first(&vm->va, 0, ~0ULL);
		while (node) {
			mapping = container_of((struct rb_node *)node,
					       struct amdgpu_bo_va_mapping, rb);
			bo = mapping->bo_va->base.bo;

			if (!amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm,
							 start << PAGE_SHIFT,
							 last << PAGE_SHIFT,
							 &userptr)) {
				node = interval_tree_iter_next(node, 0, ~0ULL);
				continue;
			}

			pr_debug("[0x%llx 0x%llx] already userptr mapped\n",
				 start, last);
			if (bo_s && bo_l) {
				*bo_s = userptr >> PAGE_SHIFT;
				*bo_l = *bo_s + bo->tbo.ttm->num_pages - 1;
			}
			amdgpu_bo_unreserve(vm->root.bo);
			return -EADDRINUSE;
		}
		amdgpu_bo_unreserve(vm->root.bo);
	}
	return 0;
}

static struct
svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev,
						struct kfd_process *p,
						struct mm_struct *mm,
						int64_t addr)
{
	struct svm_range *prange = NULL;
	unsigned long start, last;
	uint32_t gpuid, gpuidx;
	bool is_heap_stack;
	uint64_t bo_s = 0;
	uint64_t bo_l = 0;
	int r;

	if (svm_range_get_range_boundaries(p, addr, &start, &last,
					   &is_heap_stack))
		return NULL;

	r = svm_range_check_vm(p, start, last, &bo_s, &bo_l);
	if (r != -EADDRINUSE)
		r = svm_range_check_vm_userptr(p, start, last, &bo_s, &bo_l);

	if (r == -EADDRINUSE) {
		if (addr >= bo_s && addr <= bo_l)
			return NULL;

		/* Create one page svm range if 2MB range overlapping */
		start = addr;
		last = addr;
	}

	prange = svm_range_new(&p->svms, start, last);
	if (!prange) {
		pr_debug("Failed to create prange in address [0x%llx]\n", addr);
		return NULL;
	}
	if (kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx)) {
		pr_debug("failed to get gpuid from kgd\n");
		svm_range_free(prange);
		return NULL;
	}

	if (is_heap_stack)
		prange->preferred_loc = KFD_IOCTL_SVM_LOCATION_SYSMEM;

	svm_range_add_to_svms(prange);
	svm_range_add_notifier_locked(mm, prange);

	return prange;
}

/* svm_range_skip_recover - decide if prange can be recovered
 * @prange: svm range structure
 *
 * GPU vm retry fault handle skip recover the range for cases:
 * 1. prange is on deferred list to be removed after unmap, it is stale fault,
 *    deferred list work will drain the stale fault before free the prange.
 * 2. prange is on deferred list to add interval notifier after split, or
 * 3. prange is child range, it is split from parent prange, recover later
 *    after interval notifier is added.
 *
 * Return: true to skip recover, false to recover
 */
static bool svm_range_skip_recover(struct svm_range *prange)
{
	struct svm_range_list *svms = prange->svms;

	spin_lock(&svms->deferred_list_lock);
	if (list_empty(&prange->deferred_list) &&
	    list_empty(&prange->child_list)) {
		spin_unlock(&svms->deferred_list_lock);
		return false;
	}
	spin_unlock(&svms->deferred_list_lock);

	if (prange->work_item.op == SVM_OP_UNMAP_RANGE) {
		pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] unmapped\n",
			 svms, prange, prange->start, prange->last);
		return true;
	}
	if (prange->work_item.op == SVM_OP_ADD_RANGE_AND_MAP ||
	    prange->work_item.op == SVM_OP_ADD_RANGE) {
		pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] not added yet\n",
			 svms, prange, prange->start, prange->last);
		return true;
	}
	return false;
}

static void
svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
		      int32_t gpuidx)
{
	struct kfd_process_device *pdd;

	/* fault is on different page of same range
	 * or fault is skipped to recover later
	 * or fault is on invalid virtual address
	 */
	if (gpuidx == MAX_GPU_INSTANCE) {
		uint32_t gpuid;
		int r;

		r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx);
		if (r < 0)
			return;
	}

	/* fault is recovered
	 * or fault cannot recover because GPU no access on the range
	 */
	pdd = kfd_process_device_from_gpuidx(p, gpuidx);
	if (pdd)
		WRITE_ONCE(pdd->faults, pdd->faults + 1);
}

static bool
svm_fault_allowed(struct vm_area_struct *vma, bool write_fault)
{
	unsigned long requested = VM_READ;

	if (write_fault)
		requested |= VM_WRITE;

	pr_debug("requested 0x%lx, vma permission flags 0x%lx\n", requested,
		vma->vm_flags);
	return (vma->vm_flags & requested) == requested;
}

int
svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
			uint64_t addr, bool write_fault)
{
	struct mm_struct *mm = NULL;
	struct svm_range_list *svms;
	struct svm_range *prange;
	struct kfd_process *p;
	uint64_t timestamp;
	int32_t best_loc;
	int32_t gpuidx = MAX_GPU_INSTANCE;
	bool write_locked = false;
	struct vm_area_struct *vma;
	int r = 0;

	if (!KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev)) {
		pr_debug("device does not support SVM\n");
		return -EFAULT;
	}

	p = kfd_lookup_process_by_pasid(pasid);
	if (!p) {
		pr_debug("kfd process not founded pasid 0x%x\n", pasid);
		return 0;
	}
	if (!p->xnack_enabled) {
		pr_debug("XNACK not enabled for pasid 0x%x\n", pasid);
		r = -EFAULT;
		goto out;
	}
	svms = &p->svms;

	pr_debug("restoring svms 0x%p fault address 0x%llx\n", svms, addr);

	if (atomic_read(&svms->drain_pagefaults)) {
		pr_debug("draining retry fault, drop fault 0x%llx\n", addr);
		r = 0;
		goto out;
	}

	/* p->lead_thread is available as kfd_process_wq_release flush the work
	 * before releasing task ref.
	 */
	mm = get_task_mm(p->lead_thread);
	if (!mm) {
		pr_debug("svms 0x%p failed to get mm\n", svms);
		r = 0;
		goto out;
	}

	mmap_read_lock(mm);
retry_write_locked:
	mutex_lock(&svms->lock);
	prange = svm_range_from_addr(svms, addr, NULL);
	if (!prange) {
		pr_debug("failed to find prange svms 0x%p address [0x%llx]\n",
			 svms, addr);
		if (!write_locked) {
			/* Need the write lock to create new range with MMU notifier.
			 * Also flush pending deferred work to make sure the interval
			 * tree is up to date before we add a new range
			 */
			mutex_unlock(&svms->lock);
			mmap_read_unlock(mm);
			mmap_write_lock(mm);
			write_locked = true;
			goto retry_write_locked;
		}
		prange = svm_range_create_unregistered_range(adev, p, mm, addr);
		if (!prange) {
			pr_debug("failed to create unregistered range svms 0x%p address [0x%llx]\n",
				 svms, addr);
			mmap_write_downgrade(mm);
			r = -EFAULT;
			goto out_unlock_svms;
		}
	}
	if (write_locked)
		mmap_write_downgrade(mm);

	mutex_lock(&prange->migrate_mutex);

	if (svm_range_skip_recover(prange)) {
		amdgpu_gmc_filter_faults_remove(adev, addr, pasid);
		r = 0;
		goto out_unlock_range;
	}

	timestamp = ktime_to_us(ktime_get()) - prange->validate_timestamp;
	/* skip duplicate vm fault on different pages of same range */
	if (timestamp < AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING) {
		pr_debug("svms 0x%p [0x%lx %lx] already restored\n",
			 svms, prange->start, prange->last);
		r = 0;
		goto out_unlock_range;
	}

	/* __do_munmap removed VMA, return success as we are handling stale
	 * retry fault.
	 */
	vma = find_vma(mm, addr << PAGE_SHIFT);
	if (!vma || (addr << PAGE_SHIFT) < vma->vm_start) {
		pr_debug("address 0x%llx VMA is removed\n", addr);
		r = 0;
		goto out_unlock_range;
	}

	if (!svm_fault_allowed(vma, write_fault)) {
		pr_debug("fault addr 0x%llx no %s permission\n", addr,
			write_fault ? "write" : "read");
		r = -EPERM;
		goto out_unlock_range;
	}

	best_loc = svm_range_best_restore_location(prange, adev, &gpuidx);
	if (best_loc == -1) {
		pr_debug("svms %p failed get best restore loc [0x%lx 0x%lx]\n",
			 svms, prange->start, prange->last);
		r = -EACCES;
		goto out_unlock_range;
	}

	pr_debug("svms %p [0x%lx 0x%lx] best restore 0x%x, actual loc 0x%x\n",
		 svms, prange->start, prange->last, best_loc,
		 prange->actual_loc);

	if (prange->actual_loc != best_loc) {
		if (best_loc) {
			r = svm_migrate_to_vram(prange, best_loc, mm);
			if (r) {
				pr_debug("svm_migrate_to_vram failed (%d) at %llx, falling back to system memory\n",
					 r, addr);
				/* Fallback to system memory if migration to
				 * VRAM failed
				 */
				if (prange->actual_loc)
					r = svm_migrate_vram_to_ram(prange, mm);
				else
					r = 0;
			}
		} else {
			r = svm_migrate_vram_to_ram(prange, mm);
		}
		if (r) {
			pr_debug("failed %d to migrate svms %p [0x%lx 0x%lx]\n",
				 r, svms, prange->start, prange->last);
			goto out_unlock_range;
		}
	}

	r = svm_range_validate_and_map(mm, prange, gpuidx, false, false);
	if (r)
		pr_debug("failed %d to map svms 0x%p [0x%lx 0x%lx] to gpus\n",
			 r, svms, prange->start, prange->last);

out_unlock_range:
	mutex_unlock(&prange->migrate_mutex);
out_unlock_svms:
	mutex_unlock(&svms->lock);
	mmap_read_unlock(mm);

	svm_range_count_fault(adev, p, gpuidx);

	mmput(mm);
out:
	kfd_unref_process(p);

	if (r == -EAGAIN) {
		pr_debug("recover vm fault later\n");
		amdgpu_gmc_filter_faults_remove(adev, addr, pasid);
		r = 0;
	}
	return r;
}

void svm_range_list_fini(struct kfd_process *p)
{
	struct svm_range *prange;
	struct svm_range *next;

	pr_debug("pasid 0x%x svms 0x%p\n", p->pasid, &p->svms);

	/* Ensure list work is finished before process is destroyed */
	flush_work(&p->svms.deferred_list_work);

	/*
	 * Ensure no retry fault comes in afterwards, as page fault handler will
	 * not find kfd process and take mm lock to recover fault.
	 */
	atomic_inc(&p->svms.drain_pagefaults);
	svm_range_drain_retry_fault(&p->svms);


	list_for_each_entry_safe(prange, next, &p->svms.list, list) {
		svm_range_unlink(prange);
		svm_range_remove_notifier(prange);
		svm_range_free(prange);
	}

	mutex_destroy(&p->svms.lock);

	pr_debug("pasid 0x%x svms 0x%p done\n", p->pasid, &p->svms);
}

int svm_range_list_init(struct kfd_process *p)
{
	struct svm_range_list *svms = &p->svms;
	int i;

	svms->objects = RB_ROOT_CACHED;
	mutex_init(&svms->lock);
	INIT_LIST_HEAD(&svms->list);
	atomic_set(&svms->evicted_ranges, 0);
	atomic_set(&svms->drain_pagefaults, 0);
	INIT_DELAYED_WORK(&svms->restore_work, svm_range_restore_work);
	INIT_WORK(&svms->deferred_list_work, svm_range_deferred_list_work);
	INIT_LIST_HEAD(&svms->deferred_range_list);
	spin_lock_init(&svms->deferred_list_lock);

	for (i = 0; i < p->n_pdds; i++)
		if (KFD_IS_SVM_API_SUPPORTED(p->pdds[i]->dev))
			bitmap_set(svms->bitmap_supported, i, 1);

	return 0;
}

/**
 * svm_range_check_vm - check if virtual address range mapped already
 * @p: current kfd_process
 * @start: range start address, in pages
 * @last: range last address, in pages
 * @bo_s: mapping start address in pages if address range already mapped
 * @bo_l: mapping last address in pages if address range already mapped
 *
 * The purpose is to avoid virtual address ranges already allocated by
 * kfd_ioctl_alloc_memory_of_gpu ioctl.
 * It looks for each pdd in the kfd_process.
 *
 * Context: Process context
 *
 * Return 0 - OK, if the range is not mapped.
 * Otherwise error code:
 * -EADDRINUSE - if address is mapped already by kfd_ioctl_alloc_memory_of_gpu
 * -ERESTARTSYS - A wait for the buffer to become unreserved was interrupted by
 * a signal. Release all buffer reservations and return to user-space.
 */
static int
svm_range_check_vm(struct kfd_process *p, uint64_t start, uint64_t last,
		   uint64_t *bo_s, uint64_t *bo_l)
{
	struct amdgpu_bo_va_mapping *mapping;
	struct interval_tree_node *node;
	uint32_t i;
	int r;

	for (i = 0; i < p->n_pdds; i++) {
		struct amdgpu_vm *vm;

		if (!p->pdds[i]->drm_priv)
			continue;

		vm = drm_priv_to_vm(p->pdds[i]->drm_priv);
		r = amdgpu_bo_reserve(vm->root.bo, false);
		if (r)
			return r;

		node = interval_tree_iter_first(&vm->va, start, last);
		if (node) {
			pr_debug("range [0x%llx 0x%llx] already TTM mapped\n",
				 start, last);
			mapping = container_of((struct rb_node *)node,
					       struct amdgpu_bo_va_mapping, rb);
			if (bo_s && bo_l) {
				*bo_s = mapping->start;
				*bo_l = mapping->last;
			}
			amdgpu_bo_unreserve(vm->root.bo);
			return -EADDRINUSE;
		}
		amdgpu_bo_unreserve(vm->root.bo);
	}

	return 0;
}

/**
 * svm_range_is_valid - check if virtual address range is valid
 * @p: current kfd_process
 * @start: range start address, in pages
 * @size: range size, in pages
 *
 * Valid virtual address range means it belongs to one or more VMAs
 *
 * Context: Process context
 *
 * Return:
 *  0 - OK, otherwise error code
 */
static int
svm_range_is_valid(struct kfd_process *p, uint64_t start, uint64_t size)
{
	const unsigned long device_vma = VM_IO | VM_PFNMAP | VM_MIXEDMAP;
	struct vm_area_struct *vma;
	unsigned long end;
	unsigned long start_unchg = start;

	start <<= PAGE_SHIFT;
	end = start + (size << PAGE_SHIFT);
	do {
		vma = find_vma(p->mm, start);
		if (!vma || start < vma->vm_start ||
		    (vma->vm_flags & device_vma))
			return -EFAULT;
		start = min(end, vma->vm_end);
	} while (start < end);

	return svm_range_check_vm(p, start_unchg, (end - 1) >> PAGE_SHIFT, NULL,
				  NULL);
}

/**
 * svm_range_best_prefetch_location - decide the best prefetch location
 * @prange: svm range structure
 *
 * For xnack off:
 * If range map to single GPU, the best prefetch location is prefetch_loc, which
 * can be CPU or GPU.
 *
 * If range is ACCESS or ACCESS_IN_PLACE by mGPUs, only if mGPU connection on
 * XGMI same hive, the best prefetch location is prefetch_loc GPU, othervise
 * the best prefetch location is always CPU, because GPU can not have coherent
 * mapping VRAM of other GPUs even with large-BAR PCIe connection.
 *
 * For xnack on:
 * If range is not ACCESS_IN_PLACE by mGPUs, the best prefetch location is
 * prefetch_loc, other GPU access will generate vm fault and trigger migration.
 *
 * If range is ACCESS_IN_PLACE by mGPUs, only if mGPU connection on XGMI same
 * hive, the best prefetch location is prefetch_loc GPU, otherwise the best
 * prefetch location is always CPU.
 *
 * Context: Process context
 *
 * Return:
 * 0 for CPU or GPU id
 */
static uint32_t
svm_range_best_prefetch_location(struct svm_range *prange)
{
	DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE);
	uint32_t best_loc = prange->prefetch_loc;
	struct kfd_process_device *pdd;
	struct amdgpu_device *bo_adev;
	struct kfd_process *p;
	uint32_t gpuidx;

	p = container_of(prange->svms, struct kfd_process, svms);

	if (!best_loc || best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED)
		goto out;

	bo_adev = svm_range_get_adev_by_id(prange, best_loc);
	if (!bo_adev) {
		WARN_ONCE(1, "failed to get device by id 0x%x\n", best_loc);
		best_loc = 0;
		goto out;
	}

	if (p->xnack_enabled)
		bitmap_copy(bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);
	else
		bitmap_or(bitmap, prange->bitmap_access, prange->bitmap_aip,
			  MAX_GPU_INSTANCE);

	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
		if (!pdd) {
			pr_debug("failed to get device by idx 0x%x\n", gpuidx);
			continue;
		}

		if (pdd->dev->adev == bo_adev)
			continue;

		if (!amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) {
			best_loc = 0;
			break;
		}
	}

out:
	pr_debug("xnack %d svms 0x%p [0x%lx 0x%lx] best loc 0x%x\n",
		 p->xnack_enabled, &p->svms, prange->start, prange->last,
		 best_loc);

	return best_loc;
}

/* FIXME: This is a workaround for page locking bug when some pages are
 * invalid during migration to VRAM
 */
void svm_range_prefault(struct svm_range *prange, struct mm_struct *mm,
			void *owner)
{
	struct hmm_range *hmm_range;
	int r;

	if (prange->validated_once)
		return;

	r = amdgpu_hmm_range_get_pages(&prange->notifier, mm, NULL,
				       prange->start << PAGE_SHIFT,
				       prange->npages, &hmm_range,
				       false, true, owner);
	if (!r) {
		amdgpu_hmm_range_get_pages_done(hmm_range);
		prange->validated_once = true;
	}
}

/* svm_range_trigger_migration - start page migration if prefetch loc changed
 * @mm: current process mm_struct
 * @prange: svm range structure
 * @migrated: output, true if migration is triggered
 *
 * If range perfetch_loc is GPU, actual loc is cpu 0, then migrate the range
 * from ram to vram.
 * If range prefetch_loc is cpu 0, actual loc is GPU, then migrate the range
 * from vram to ram.
 *
 * If GPU vm fault retry is not enabled, migration interact with MMU notifier
 * and restore work:
 * 1. migrate_vma_setup invalidate pages, MMU notifier callback svm_range_evict
 *    stops all queues, schedule restore work
 * 2. svm_range_restore_work wait for migration is done by
 *    a. svm_range_validate_vram takes prange->migrate_mutex
 *    b. svm_range_validate_ram HMM get pages wait for CPU fault handle returns
 * 3. restore work update mappings of GPU, resume all queues.
 *
 * Context: Process context
 *
 * Return:
 * 0 - OK, otherwise - error code of migration
 */
static int
svm_range_trigger_migration(struct mm_struct *mm, struct svm_range *prange,
			    bool *migrated)
{
	uint32_t best_loc;
	int r = 0;

	*migrated = false;
	best_loc = svm_range_best_prefetch_location(prange);

	if (best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED ||
	    best_loc == prange->actual_loc)
		return 0;

	if (!best_loc) {
		r = svm_migrate_vram_to_ram(prange, mm);
		*migrated = !r;
		return r;
	}

	r = svm_migrate_to_vram(prange, best_loc, mm);
	*migrated = !r;

	return r;
}

int svm_range_schedule_evict_svm_bo(struct amdgpu_amdkfd_fence *fence)
{
	if (!fence)
		return -EINVAL;

	if (dma_fence_is_signaled(&fence->base))
		return 0;

	if (fence->svm_bo) {
		WRITE_ONCE(fence->svm_bo->evicting, 1);
		schedule_work(&fence->svm_bo->eviction_work);
	}

	return 0;
}

static void svm_range_evict_svm_bo_worker(struct work_struct *work)
{
	struct svm_range_bo *svm_bo;
	struct kfd_process *p;
	struct mm_struct *mm;

	svm_bo = container_of(work, struct svm_range_bo, eviction_work);
	if (!svm_bo_ref_unless_zero(svm_bo))
		return; /* svm_bo was freed while eviction was pending */

	/* svm_range_bo_release destroys this worker thread. So during
	 * the lifetime of this thread, kfd_process and mm will be valid.
	 */
	p = container_of(svm_bo->svms, struct kfd_process, svms);
	mm = p->mm;
	if (!mm)
		return;

	mmap_read_lock(mm);
	spin_lock(&svm_bo->list_lock);
	while (!list_empty(&svm_bo->range_list)) {
		struct svm_range *prange =
				list_first_entry(&svm_bo->range_list,
						struct svm_range, svm_bo_list);
		int retries = 3;

		list_del_init(&prange->svm_bo_list);
		spin_unlock(&svm_bo->list_lock);

		pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms,
			 prange->start, prange->last);

		mutex_lock(&prange->migrate_mutex);
		do {
			svm_migrate_vram_to_ram(prange,
						svm_bo->eviction_fence->mm);
		} while (prange->actual_loc && --retries);
		WARN(prange->actual_loc, "Migration failed during eviction");

		mutex_lock(&prange->lock);
		prange->svm_bo = NULL;
		mutex_unlock(&prange->lock);

		mutex_unlock(&prange->migrate_mutex);

		spin_lock(&svm_bo->list_lock);
	}
	spin_unlock(&svm_bo->list_lock);
	mmap_read_unlock(mm);

	dma_fence_signal(&svm_bo->eviction_fence->base);
	/* This is the last reference to svm_bo, after svm_range_vram_node_free
	 * has been called in svm_migrate_vram_to_ram
	 */
	WARN_ONCE(kref_read(&svm_bo->kref) != 1, "This was not the last reference\n");
	svm_range_bo_unref(svm_bo);
}

static int
svm_range_set_attr(struct kfd_process *p, uint64_t start, uint64_t size,
		   uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs)
{
	struct mm_struct *mm = current->mm;
	struct list_head update_list;
	struct list_head insert_list;
	struct list_head remove_list;
	struct svm_range_list *svms;
	struct svm_range *prange;
	struct svm_range *next;
	int r = 0;

	pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] pages 0x%llx\n",
		 p->pasid, &p->svms, start, start + size - 1, size);

	r = svm_range_check_attr(p, nattr, attrs);
	if (r)
		return r;

	svms = &p->svms;

	svm_range_list_lock_and_flush_work(svms, mm);

	r = svm_range_is_valid(p, start, size);
	if (r) {
		pr_debug("invalid range r=%d\n", r);
		mmap_write_unlock(mm);
		goto out;
	}

	mutex_lock(&svms->lock);

	/* Add new range and split existing ranges as needed */
	r = svm_range_add(p, start, size, nattr, attrs, &update_list,
			  &insert_list, &remove_list);
	if (r) {
		mutex_unlock(&svms->lock);
		mmap_write_unlock(mm);
		goto out;
	}
	/* Apply changes as a transaction */
	list_for_each_entry_safe(prange, next, &insert_list, list) {
		svm_range_add_to_svms(prange);
		svm_range_add_notifier_locked(mm, prange);
	}
	list_for_each_entry(prange, &update_list, update_list) {
		svm_range_apply_attrs(p, prange, nattr, attrs);
		/* TODO: unmap ranges from GPU that lost access */
	}
	list_for_each_entry_safe(prange, next, &remove_list, update_list) {
		pr_debug("unlink old 0x%p prange 0x%p [0x%lx 0x%lx]\n",
			 prange->svms, prange, prange->start,
			 prange->last);
		svm_range_unlink(prange);
		svm_range_remove_notifier(prange);
		svm_range_free(prange);
	}

	mmap_write_downgrade(mm);
	/* Trigger migrations and revalidate and map to GPUs as needed. If
	 * this fails we may be left with partially completed actions. There
	 * is no clean way of rolling back to the previous state in such a
	 * case because the rollback wouldn't be guaranteed to work either.
	 */
	list_for_each_entry(prange, &update_list, update_list) {
		bool migrated;

		mutex_lock(&prange->migrate_mutex);

		r = svm_range_trigger_migration(mm, prange, &migrated);
		if (r)
			goto out_unlock_range;

		if (migrated && !p->xnack_enabled) {
			pr_debug("restore_work will update mappings of GPUs\n");
			mutex_unlock(&prange->migrate_mutex);
			continue;
		}

		r = svm_range_validate_and_map(mm, prange, MAX_GPU_INSTANCE,
					       true, true);
		if (r)
			pr_debug("failed %d to map svm range\n", r);

out_unlock_range:
		mutex_unlock(&prange->migrate_mutex);
		if (r)
			break;
	}

	svm_range_debug_dump(svms);

	mutex_unlock(&svms->lock);
	mmap_read_unlock(mm);
out:
	pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] done, r=%d\n", p->pasid,
		 &p->svms, start, start + size - 1, r);

	return r;
}

static int
svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size,
		   uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs)
{
	DECLARE_BITMAP(bitmap_access, MAX_GPU_INSTANCE);
	DECLARE_BITMAP(bitmap_aip, MAX_GPU_INSTANCE);
	bool get_preferred_loc = false;
	bool get_prefetch_loc = false;
	bool get_granularity = false;
	bool get_accessible = false;
	bool get_flags = false;
	uint64_t last = start + size - 1UL;
	struct mm_struct *mm = current->mm;
	uint8_t granularity = 0xff;
	struct interval_tree_node *node;
	struct svm_range_list *svms;
	struct svm_range *prange;
	uint32_t prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
	uint32_t location = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
	uint32_t flags_and = 0xffffffff;
	uint32_t flags_or = 0;
	int gpuidx;
	uint32_t i;
	int r = 0;

	pr_debug("svms 0x%p [0x%llx 0x%llx] nattr 0x%x\n", &p->svms, start,
		 start + size - 1, nattr);

	/* Flush pending deferred work to avoid racing with deferred actions from
	 * previous memory map changes (e.g. munmap). Concurrent memory map changes
	 * can still race with get_attr because we don't hold the mmap lock. But that
	 * would be a race condition in the application anyway, and undefined
	 * behaviour is acceptable in that case.
	 */
	flush_work(&p->svms.deferred_list_work);

	mmap_read_lock(mm);
	r = svm_range_is_valid(p, start, size);
	mmap_read_unlock(mm);
	if (r) {
		pr_debug("invalid range r=%d\n", r);
		return r;
	}

	for (i = 0; i < nattr; i++) {
		switch (attrs[i].type) {
		case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC:
			get_preferred_loc = true;
			break;
		case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC:
			get_prefetch_loc = true;
			break;
		case KFD_IOCTL_SVM_ATTR_ACCESS:
			get_accessible = true;
			break;
		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
			get_flags = true;
			break;
		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
			get_granularity = true;
			break;
		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
			fallthrough;
		default:
			pr_debug("get invalid attr type 0x%x\n", attrs[i].type);
			return -EINVAL;
		}
	}

	svms = &p->svms;

	mutex_lock(&svms->lock);

	node = interval_tree_iter_first(&svms->objects, start, last);
	if (!node) {
		pr_debug("range attrs not found return default values\n");
		svm_range_set_default_attributes(&location, &prefetch_loc,
						 &granularity, &flags_and);
		flags_or = flags_and;
		if (p->xnack_enabled)
			bitmap_copy(bitmap_access, svms->bitmap_supported,
				    MAX_GPU_INSTANCE);
		else
			bitmap_zero(bitmap_access, MAX_GPU_INSTANCE);
		bitmap_zero(bitmap_aip, MAX_GPU_INSTANCE);
		goto fill_values;
	}
	bitmap_copy(bitmap_access, svms->bitmap_supported, MAX_GPU_INSTANCE);
	bitmap_copy(bitmap_aip, svms->bitmap_supported, MAX_GPU_INSTANCE);

	while (node) {
		struct interval_tree_node *next;

		prange = container_of(node, struct svm_range, it_node);
		next = interval_tree_iter_next(node, start, last);

		if (get_preferred_loc) {
			if (prange->preferred_loc ==
					KFD_IOCTL_SVM_LOCATION_UNDEFINED ||
			    (location != KFD_IOCTL_SVM_LOCATION_UNDEFINED &&
			     location != prange->preferred_loc)) {
				location = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
				get_preferred_loc = false;
			} else {
				location = prange->preferred_loc;
			}
		}
		if (get_prefetch_loc) {
			if (prange->prefetch_loc ==
					KFD_IOCTL_SVM_LOCATION_UNDEFINED ||
			    (prefetch_loc != KFD_IOCTL_SVM_LOCATION_UNDEFINED &&
			     prefetch_loc != prange->prefetch_loc)) {
				prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
				get_prefetch_loc = false;
			} else {
				prefetch_loc = prange->prefetch_loc;
			}
		}
		if (get_accessible) {
			bitmap_and(bitmap_access, bitmap_access,
				   prange->bitmap_access, MAX_GPU_INSTANCE);
			bitmap_and(bitmap_aip, bitmap_aip,
				   prange->bitmap_aip, MAX_GPU_INSTANCE);
		}
		if (get_flags) {
			flags_and &= prange->flags;
			flags_or |= prange->flags;
		}

		if (get_granularity && prange->granularity < granularity)
			granularity = prange->granularity;

		node = next;
	}
fill_values:
	mutex_unlock(&svms->lock);

	for (i = 0; i < nattr; i++) {
		switch (attrs[i].type) {
		case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC:
			attrs[i].value = location;
			break;
		case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC:
			attrs[i].value = prefetch_loc;
			break;
		case KFD_IOCTL_SVM_ATTR_ACCESS:
			gpuidx = kfd_process_gpuidx_from_gpuid(p,
							       attrs[i].value);
			if (gpuidx < 0) {
				pr_debug("invalid gpuid %x\n", attrs[i].value);
				return -EINVAL;
			}
			if (test_bit(gpuidx, bitmap_access))
				attrs[i].type = KFD_IOCTL_SVM_ATTR_ACCESS;
			else if (test_bit(gpuidx, bitmap_aip))
				attrs[i].type =
					KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE;
			else
				attrs[i].type = KFD_IOCTL_SVM_ATTR_NO_ACCESS;
			break;
		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
			attrs[i].value = flags_and;
			break;
		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
			attrs[i].value = ~flags_or;
			break;
		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
			attrs[i].value = (uint32_t)granularity;
			break;
		}
	}

	return 0;
}

int
svm_ioctl(struct kfd_process *p, enum kfd_ioctl_svm_op op, uint64_t start,
	  uint64_t size, uint32_t nattrs, struct kfd_ioctl_svm_attribute *attrs)
{
	int r;

	start >>= PAGE_SHIFT;
	size >>= PAGE_SHIFT;

	switch (op) {
	case KFD_IOCTL_SVM_OP_SET_ATTR:
		r = svm_range_set_attr(p, start, size, nattrs, attrs);
		break;
	case KFD_IOCTL_SVM_OP_GET_ATTR:
		r = svm_range_get_attr(p, start, size, nattrs, attrs);
		break;
	default:
		r = EINVAL;
		break;
	}

	return r;
}